Main

Glutathione S-transferases

The Glutathione S-transferases (GSTs) comprise a family of phase II detoxifying enzymes that catalyze a large number of reactions taking place between the cytosolic glutathione and compounds containing an electrophilic center.1 These enzymes are involved in the elimination of xenobiotics and endogenous products of oxidative stress formed as a result of aerobic metabolism, exposure to ionizing radiation or any other process that causes cellular damage. Substrates for GSTs include acetaldehyde and several polyaromatic hydrocarbons (PAHs) found in tobacco smoke. The main steps for GST catalysis includes the formation of a complex with the cytosolic glutathione and the ionization of the sulfydryl group of this enzyme bound to glutathione to yield a highly reactive thiolate anion through hydrogen bonding with the adjacent hydroxyl group. The enhancement of nucleophilicity activates the gluthathione and it can react with various electrophilic substrates containing carbon, nitrogen, or sulfur atoms. The result of this conjugation leads to elimination of the carcinogens from the body.

Based on sequence similarities, human cytosolic GSTs have been grouped into at least four major gene families (alpha, mu, pi, and theta). The alpha class is located in chromosome 6p12, the mu class in chromosome 1p13, pi in chromosome 11, and theta in chromosome 22. Various isoenzymes have been identified for the alpha (A1–12), mu (M1–M5), pi (P1–P2), and theta class gene families (T1–T2). The GSTM1, M2, M3, T1, and P1 are expressed in a variety of tissues including the squamous epithelium of the oral cavity2 and are involved in the detoxification of various polycyclic aromatic hydrocarbons, including benzo[a]pyrene-7,8-diol-9,10-epoxide,3 one of the most important carcinogens found in tobacco smoke, by catalyzing the conversion of the reactive electrophiles to inactive, water soluble conjugates that can be easily excreted.4 The GSTM1 isoenzyme together with the alcohol dehydrogenase is also involved in the oxidation of ethanol to acetaldehyde.5

Three alleles have been identified at the GSTM1 locus: GSTM1*0, GSTM1*A, and GSTM1*B. The GSTM1*A and GSTM1*B differ by a C→G substitution at base position 534.6 This C→G substitution results in a substitution of Lys→Asn at amino acid 172. These result in monodimers (GSTM1A-1A, GSTM1B-1B) or heterodimers (GSTM1A-1B), but in vitro studies suggest that their activities are similar.7 The GSTM1*0, also called the null allele, is a huge deletion at GSTM1 and homozygotes express no GSTM1 protein activity.8 These subjects may potentially accumulate more DNA adducts and mutagen induced damage that may cause differences in susceptibility to tumorigenesis.9

Cytochrome P450s

The cytochrome P450 family (CYP) of heme monooxygenases comprise phase I enzymes that oxidate a wide variety of endogenous and exogenous compounds using atmospheric oxygen.10 Currently, more than 270 CYP gene families are known. Humans have 57 potentially functional P450 genes and 33 pseudogenes arranged into 18 families and 42 subfamilies.11

The CYP1A1 gene belongs to the CYP1 subfamily and encodes for the enzyme aryl hydrocarbon hydrolase, which is involved in the activation of PAHs and aromatic amines12 and is expressed in oral tissue.13 Various studies show that CYP1A1 catalyzes the initial metabolism of benzo[a]pyrene.4,14 The CYP1A1 gene is located in chromosome 15, band 15q22–2415 and several important single nucleotide polymorphisms have been identified. The nomenclature of these polymorphisms is now standardized16,17 but different nomenclatures were used for several years.12 The first allele presents a single base substitution of thymine by cytosine in a noncoding region of the gene at position 3801 that creates a MspI (m1) restriction site (CYP1A1*2A). A single base substitution of adenine to guanine at position 2455 in the heme binding region of exon 7 induces an amino acid change in isoleucine to valine at codon 462 and is known as the Ile/Val or exon 7 polymorphism (Ile462 Val) or CYP1A1*2C.18 In whites, this polymorphism is in complete linkage disequilibrium with the CYP1A1 MspI (CYP1A1*2B).19 Another polymorphism in exon 7, a base substitution of cytosine by adenine at position 2453, leading to the Thr461Asn polymorphism (CYP1A1*4) has been described.20 Some CYP1A1 polymorhisms have been shown to increase microsomal catalytic activity for converting procarcinogens, including PAH and aromatic amines, but the results are inconsistent.12,21–23 It has been suggested that DNA damage may depend on the link of CYP1A1 to other polymorphisms that can affect CYP1A1 transcription levels, such as polymorphisms for promoter genes, Ah receptor genes, or other metabolic genes such as GSTM1.23,24

Oral and pharyngeal cancers and risk factors

According to the International Classification of Diseases-10th revision (ICD-10) oral and pharyngeal tumors are defined as those cancers comprising the locations C00–C14. These cancers represent an important problem worldwide, with 484,628 new cases and 262,784 deaths estimated per year.25 The highest incidence and prevalence rates are observed in Melanesia, Central Asia, and Western Europe, even though rates vary depending on the gender and cancer location.25 In men, cancers of the oral cavity are eighth in terms of incidence worldwide and they are responsible for 3% of the cancers diagnosed in this gender. Pharyngeal tumors are also common in European and Central Asian countries but the incidence rates are lower.

Mortality rates are substantially lower than incidence rates, with 2.2 deaths per 100,000 people worldwide.25 The highest values are recorded in several countries of Central and Eastern Europe and the lowest in Central America and Northern Europe. In Hungary, the mortality rate is as high as 21.2 per 100,000.

Risk factors for oral and pharyngeal cancers

Since 1988, tobacco and alcohol consumption have been recognized as independent risk factors for oral cancer. Epidemiologic studies performed in all continents have found an increased risk in smokers and a dose-response relationship with daily cigarettes and duration of habit.

An excessive consumption of alcoholic beverages has been associated with oral and pharyngeal cancer, with relative risks sometimes higher than those found for smokers.26–28 The risk associated with alcohol increases with consumption26,29–32, duration, starting age and type of alcohol beverage.26,29,33,34 When joint consumption of alcohol and tobacco was investigated, the great majority of the literature suggests that the joint effect is multiplicative or, at least, greater than additive.26,35

Human papillomavirus (HPV) is another possible key factor in the etiology of oral and pharyngeal cancers36–38; two recent studies reported a high risk of oral and pharyngeal cancer associated with HPV16 and HPV18 (odds ratio [OR]: 61 and OR: 63).31,39

Metabolic genes and risk of oral and pharyngeal cancers

CYP1A1 and GSTM1 are important enzymes in the metabolism of tobacco carcinogens, which involves a balance between the activation steps mediated by the cytochrome P450 system and the detoxification steps involving GSTM1 that catalyze the conversion of the reactive electrophiles to inactive, water soluble conjugates that can be easily removed.4

Previous systematic reviews, meta-analysis and pooled analysis, have reported a relationship between the GSTM1 null genotype and the risk of head and neck cancer2,40–43 but the only report that stratified the analysis for cancer site41 found important differences in risk for oral and laryngeal tumors. No association was found for the CYP1A1 (Ile/val) polymorphism in this last assessment. Because different patterns of GST and CYP1A1 enzyme expression have been shown in oral and pharyngeal epithelium in comparison with laryngeal epithelium,12,44 we conducted a pooled and meta-analysis to evaluate the relationship between these polymorphisms and oral and pharyngeal tumors, and we explored the combined effects of polymorphisms in these two genes along with their interaction with smoking.

METHODS

Selection criteria

The association of GSTM1 and CYP1A1 with oral and pharyngeal cancers was determined by meta-analysis of publications identified in a systematic review as well as by a pooled analysis using both published and unpublished data from the Genetic Susceptibility to Environmental Carcinogens (GSEC) database. A bibliographic search was carried out in the MEDLINE and EMBASE databases to identify studies on oral and pharyngeal cancers published up to October 17, 2007. The search strategy used was: (oral or buccal or mouth or “head and neck” or pharyngeal or pharynx or oropharyngeal) and (cancer or neoplasms or tumor* or tumour* or carcinoma* or carcinogenesis) and (“glutathione transferase” or “glutathione S transferase” or “glutathione S-transferase” or GSTM1 or “cytochrome P450 enzyme system” or “cytochrome P450 CYP1A1” or CYP1A1). A manual review of the bibliographic references cited in the selected articles was undertaken to retrieve articles that might have been missed in the search. Articles were independently reviewed by two researchers and the inclusion/exclusion was made by consensus on the basis of pre-established selection criteria. The inclusion criteria were: (1) articles published in English, Spanish, Italian, or French, and (2) studies that assessed the association between the polymorphisms of the genes under study and oral and pharyngeal cancers. The exclusion criteria were: (1) studies that included only cases; (2) studies that assessed the risk of secondary tumors, recurrence, or response to treatment; (3) studies where patients were overlapped; and (4) studies that included nasopharyngeal cases. When several studies included the same population we included only the most updated one.

The meta-analysis included only those articles that provided results that allowed for the calculation of crude risks for oral and/or pharyngeal tumors. Crude ORs were used to obtain comparable estimates across studies. For each study included the author, year of publication, country where the study was carried out, number, race, and gender of patients and controls, control source (hospital based or population based), tumor site, and matching of cases and controls were rigorously tabulated. The bibliographic search led to the identification of 56 original articles. Of these, five did not include data on the genes involved in this analysis,45–49 three did not provide the data that was needed to calculate the ORs for oral and pharyngeal sites,50–52 therefore were not further evaluated. Of the remaining 48 articles, 18 were excluded from the meta-analysis because they did not provide head and neck subsite specific data and subjects with laryngeal tumors were not distinguished from the oral/pharyngeal group.53–70 Thirty publications were used in the meta-analysis including a total of 7783 subjects (3177 cases and 4606 controls). They were all case-control studies. There were two studies with overlapping subjects but reported data separately for GSTM171 and CYP1A1.72 Two other studies each reported separately data for CYP1A1 Msp173 and exon 7.74 However, both publications reported overlapping data for GSTM1. Therefore, there were 26 studies with results on GSTM1 deletion, 11 on CYP1A1 Ile/Val polymorphism, 6 on CYP1A1 MspI polymorphism. Only three studies assessed the combined GSTM1/CYP1A1 MspI polymorphisms and one the GSTM1/CYP1A1 exon 7 polymorphisms.

Data collection

The pooled analysis was performed using information from the GSEC database (http://www.upci.upmc.edu/research/ccps/ccontrol/g_intro.html).75,76 Briefly, the GSEC study is a collection of data from both published and unpublished case-control studies of metabolic gene polymorphisms and cancer. All of the investigators of the published studies for which the GSEC database did not contain their data were contacted and invited to provide their data for this specific pooled analysis. The investigators for the other studies that were excluded because of insufficient head and neck subsite specific data were also contacted. Of the 30 studies in the meta-analysis, data for 14 studies were obtained for GSTM1 and/or CYP1A1.5,44,71,72,77–86 However, two of these studies reported CYP1A1 and GSTM1 data separately for the same subjects and were counted as a single study.71,72 Among these 13 published studies, three provided unpublished data for CYP1A1 polymorphisms for the same subjects. The GSEC database also had one study with unpublished data for GSTM1 deletion (Foulkes et al., unpublished data) and another study with unpublished data for both GSTM1 and CYP1A1 (Ruano-Ravina et al., unpublished data). There were also seven additional published studies that were previously excluded from the meta-analysis, which were now included in the pooled analysis because the raw data allowed us to define specific head and neck subsites.57,59,65,66,69,70,87 Although there were 22 studies available, 2 of them reported overlapping data for GSTM1. Therefore, the pooled analysis included 21 datasets, with 9397 subjects (3130 cases and 6267 controls).

Statistical analysis

All statistical analyses were carried out using STATA SE (version 10) software (StataCorp LP, College Station, TX). For the meta-analysis, the frequency of cases and controls was extracted from each publication and study-specific crude ORs were calculated along with their 95% confidence intervals (CIs). The Q statistics were used to test for heterogeneity among the studies for GSTM1 deletion and CYP1A1 polymorphisms. When heterogeneity was observed a random-effects model was used to calculate the summary ORs for the combined studies, when heterogeneity was not observed a fixed-effects model was used. Publication bias was determined by performing the Eggers test. To explore the between study heterogeneity, sensitivity analyses were performed, to identify the influence of the individual studies on the combined OR. When a study was identified, the analysis was repeated excluding such study to assess if homogeneity between the remaining studies was reached.

In the pooled analysis for each gene, crude ORs for their overall association with oral/pharyngeal cancer were calculated. ORs adjusted for potential confounders were calculated using multivariable logistic regression models. Crude and adjusted ORs were also calculated for each gene, stratifying by control source (healthy versus hospital), smoking status, race and tumors site (oral cavity versus pharynx). The Mantel-Haenszel test was used to assess differences between stratum-specific ORs.

From south east/south Asia publications, three of the five available studies included data on consumption of other tobacco or had tobacco chewing habits; these patients were included in the pooled analysis, but the data on other tobacco was not analyzed for the present publication. Smoking status was defined as never and ex, current, or ever smokers. All smoking data were recoded into a standardized variable: ever/never smoking. Patients were classified as never smokers if they smoked < 100 cigarettes in their lifetime, and ever smokers if defined by the individual studies either as ex, current, or ever (current and ex) smokers.

RESULTS

Of the 30 studies included in the meta-analysis, 17 were carried out in Asian countries,73,74,77,78,80,84,86–96 seven in American countries,71,72,79,83,85,97,98 and six in Europe.5,44,81,82,99,100 Hospital patients were used as controls in 16 studies.5,44,71,72,78,81–85,88,92,93,97–99 The number of cases in the studies included in the meta-analysis for GSTM1 deletion varied from 21 to 451 patients. All studies undertaken in Europe included <150 cases, with two of these having <50 cases.5,82 For the CYP1A1 analysis, the case numbers ranged from 45 to 446 subjects.

Population frequencies

The frequency of the GSTM1 null in the control group ranged from 24% to 58.9%, with considerable variation depending on the area the study was carried out. In Asia, large differences could be observed between countries. The frequencies in India varied from 24% to 37%,78,80,84,90,96 in Japan from 39.8% to 48.773,77,91–94, although the only study from Taiwan observed a frequency of 57.7%.95 In South America these values ranged from 38.2% to 48.771,79,97,98 and in Europe and United States from 51% to 54.8%.5,44,81–83,99,100

For the CYP1A1 exon 7 polymorphism, large geographical heterogeneity could be observed. The frequency of the homozygote genotype for the variant allele in the controls was absent or very low in Europe (0–6%) whereas the heterozygous genotype was very rare (6–9.3%).44,100 In Asia, the heterozygous genotype was present in 32.4–53.4% of the control subjects.74,87,88,93 In Brazil and Puerto Rico this polymorphism was found in 19–30% of the subjects.72,79,85 The combined frequency of the homozygous and heterozygous genotype of the variant allele for the single study in the United States was 7.4%.83 The CYP1A1 MspI heterozygous variant allele (m1/m2) was present in 30–59.5% of the Asian control population.73,84,86,88,92 The only European study that assessed this polymorphism reported a frequency of 9.3% for the variant allele.44 The homozygous allele was very rare in all populations (1–10.6%).

Meta-analysis

The overall meta-OR for GSTM1 null was not reported because of the large heterogeneity between studies (Q test P value < 0.001; data not shown). We performed a sensitivity analysis and identified one study that appeared to influence the overall meta-OR,80 however, heterogeneity was still observed after exclusion of this study. In an effort to further explore the observed heterogeneity, we stratified the studies by race. The study-specific and meta-ORs for GSTM1 are shown for whites, Asians, and others (i.e., studies that did not specify ethnicity or included more than one ethnic group) in Table 1. There was no increased risk of oral and pharyngeal cancer with the GSTM1 deletion among whites (OR: 1.1, 95% CI: 0.9–1.3), and no evidence of publication bias (Eggers test P value = 0.19). For Asians and all other ethnic groups and studies with mixed populations, there was still large heterogeneity between studies (Q test, P value < 0.001); therefore, the overall meta-OR was not reported although there was no evidence of publication bias (Eggers test P value = 0.77 for Asian studies and 0.80 for other studies). Sensitivity analysis of the Asian studies identified a data set that seemed to influence the meta-ORs. When this study was excluded, homogeneity was observed among the remaining studies (Q test, P value = 0.186). There was a statistically significant increase in the risk of oral and pharyngeal cancer with the GSTM1 deletion (OR: 1.6, 95% CI: 1.3–2.0). There was no evidence of publication bias (Eggers test P value = 0.819). For the remaining studies (i.e., studies that did not specify ethnicity or included more than one ethnic group), heterogeneity was still observed even after exclusion of the outlier,80 (Q test, P value 0.005); this was likely due to the mixed populations grouped in this category.

Table 1 Description of studies included in the meta-analysis for GSTM1
Full size table

The 15 studies with data reported on CYP1A1 MspI and/or exon7 (Ile/Val) are summarized in Table 2. There were 11 studies overall with CYP1A1 (Ile/Val) data and 7 studies with CYP1A1 MspI data. Nine studies reported data on the associations between the Ile/Val polymorphism and risk of oral and pharyngeal cancers, 6 studies reported associations for the Val/Val polymorphism, and 10 reported associations for all variants combined (i.e., Ile/Val and Val/Val). For each of these groups, the studies were statistically significantly heterogeneous (Q test, P value < 0.001), therefore no overall meta-ORs were reported. There was no evidence of publication bias (Eggers test P value: Ile/Val = 0.945, Val/Val = 0.625, and Ile/Val + Val/Val = 0.199). Sensitivity analysis of these studies identified a data set that appeared to influence the meta-ORs. However, exclusion of this study did not resolve the heterogeneity between the remaining studies. The observed heterogeneity is likely due to misclassification, because most of the earlier studies used a laboratory method that may not accurately distinguish between the exon 7 variant alleles.

Table 2 Description of the studies included in the meta analysis for CYP1A1
Full size table

Among the five studies with CYP1A1 MspI data, all except for one study reported the associations for the m1m2, m2m2, and combined variants (m1m2 + m2m2). The studies that reported data for the m1m2 and combined variants (m1m2 + m2m2) were statistically significantly heterogeneous; therefore the meta-ORs were not reported. No publication bias was observed (Eggers test P value: m1m2 = 0.389 and m1m2 + m2m2 = 0.339). There was an increased risk of oral and pharyngeal cancers for patients with the m2m2 variant (meta-OR: 1.9, 95% CI: 1.4–2.7). There was no evidence of publication bias (Eggers test P value: m2m2 = 0.595). Sensitivity analyses identified a study that influenced the meta-ORs for the m1m2 and combined variants (m1m2 + m2m2).92 After excluding this data set, homogeneity was obtained; no association for the m1m2 or combined variants and oral and pharyngeal cancer was observed (m1m2 + m2m2) (m1m2, Q test, P value = 0.625, OR: 0.9, 95% CI: 0.8–1.1, m1m2 + m2m2, Q test, P value = 0.798, OR: 1.0, 95% CI: 0.9–1.2). There was no evidence of publication bias (Eggers test P value: m1m2 = 0.628, m1m2 + m2m2 = 0.407).

Only one study evaluated the interaction between the GSTM1 null and CYP1A1 (Ile/Val) polymorphism, and three evaluated the interaction between the GSTM1 null and CYP1A1 MspI polymorphism (Table 3). The overall meta-OR for GSTM1 null + m1m1 was not reported because the studies were statistically significantly heterogeneous (Q test P value = 0.002). There seemed to be an increased risk of oral and pharyngeal cancers for the GSTM1 wt + m1m2 or m2m2 (meta-OR: 1.6, 95% CI: 1.0–2.7) and the GSTM1 null + m1m2 or m2m2 (meta-OR: 3.0, 95% CI: 1.8–5.0). However, the association was not statistically significant for all other polymorphic isoforms. There was no publication bias observed for any of these analyses.

Table 3 Description of studies included in the meta analysis for GSTM1-CYP1A1 interaction
Full size table

Pooled analysis

The GSEC pooled analysis included 21 studies (3130 cases and 6267 controls).

Significant heterogeneity was observed between the 20 studies that contained data for GSTM1. Similar to the meta-analysis, one study seemed to contribute to the heterogeneity.80 Analyses were then stratified by various covariates. There was no association between the GSTM1 deletion and oral and pharyngeal cancers (Table 4), even when the analysis was limited to healthy controls (Adjusted odds ratio [AdjOR]: 1.1, 95% CI: 0.8–1.4). A marginal statistically significant association was observed for current smokers (AdjOR: 1.2, 95% CI: 1.0–1.4) or ever smokers (AdjOR: 1.1, 95% CI: 1.0–1.3), but not in never smokers (AdjOR: 1.0, 95% CI: 0.8–1.2). The differences observed between the stratum-specific ORs for smoking were not statistically significant (P > 0.1) (data not shown). The datasets for never, ex, and current were homogeneous. (Q test, P value > 0.05) but was not for ever smokers (Q test, P value = 0.018). The GSTM1 deletion was statistically significantly associated with oral and pharyngeal cancer in African Americans and Africans (OR: 1.9, 95% CI: 1.1–3.3), but was no longer statistically significant after adjusting for confounding variables (AdjOR: 1.7, 95% CI: 0.9–3.3). There was no association between GSTM1 deletion and oral and pharyngeal cancer risk in white, Asian populations, or other ethnic groups.

Table 4 Overall and stratified odds ratios of the association of GSTM1 deletion with oral and pharyngeal cancers–pooled analysis
Full size table

The adjusted summary OR for the association of CYP1A1 MspI polymorphism and oral and pharyngeal cancers (Table 5) was not significant for the m1m2 genotype but was for the m2m2 genotype (AdjOR: 2.0, 95% CI: 1.3–3.1). Among oral and pharyngeal cancers, there was a 2-fold likelihood of having the m2m2 genotype compared with the controls in never smokers (AdjOR: 1.8, 95% CI: 1.1–2.9) but not in current or ever smokers. There was a statistically significant difference when the stratum-specific ORs for never and current smokers were compared (P value = 0.019). The association of the m2m2 variant also differed when limited to healthy controls (AdjOR- healthy controls: 1.2, 95% CI: 0.7–2.2) versus hospital controls: 1.7, 95% CI: 1.1–2.7). A statistically significant association of the m2m2 genotype was observed for white (AdjOR: 2.1, 95% CI: 1.4–3.3) but not for other ethnic groups, although these were a mixed population.

Table 5 Overall and stratified odds ratios of the association of CYP1A1 MspI polymorphism with oral and pharyngeal cancers–pooled analysis
Full size table

In contrast, there was no association between the CYP1A1 (exon7) variant and oral and pharyngeal cancers regardless of the type of controls used in the analysis (Table 6). However, there was a statistically significant association of the Val/Val genotype for ever smokers (AdjOR: 2.2, 95% CI: 1.1–4.5). Asian cases seemed to have almost a 4-fold likelihood of having the Val/Val genotype when compared with the controls; however, this was only marginally statistically significant (AdjOR: 3.5, 95% CI: 1.0–12.6).

Table 6 Overall and stratified odds ratios of the association of CYP1A1 (exon7) polymorphism with oral and pharyngeal cancers–pooled analysis
Full size table

A marginal increased risk of cancer with the GSTM1 deletion was observed when examining oral cavity (AdjOR: 1.1, 95% CI: 1.0–1.2) and pharyngeal (AdjOR: 1.3, 95% CI: 1.1–1.6) cases independently. Among subjects with oral cavity tumors, no associations were observed for CYP1A1 (exon7) but for CYP1A1 MspI polymorphisms there was a marginal association; the m2m2 genotype was significantly associated with oral cavity tumors (AdjOR: 2.0, 95% CI: 1.3–3.1) (Table 7). We were unable to determine the association of this variant genotype for subjects with pharyngeal tumors.

Table 7 The association of GSTM1, CYP1A1 polymorphisms with oral and pharyngeal cancers according to tumor site–pooled analysis
Full size table

When evaluating alcohol use, a marginal increased risk of cancer with GSTM1 deletion was observed for both never and ever drinkers (never drinkers, AdjOR: 1.2, 95% CI: 1.0–1.5, ever drinkers, AdjOR: 1.2, 95% CI: 1.0–1.3) (Table 4). There was no association of the CYP1A1 (exon7) polymorphisms with oral and pharyngeal cancer according to alcohol consumption (Table 5), but an increase risk associated with the CYP1A1 m2m2 genotype in never drinkers only was observed (AdjOR: 2.6, 95% CI: 1.5–4.3) (Table 6).

Complete GSTM1 and CYP1A1 genotype

The combination of the CYP1A1 MspI and CYP1A1 (exon7) polymorphisms was not associated with the risk of oral and pharyngeal cancers (data not shown). The combination of the GSTM1 null plus the CYP1A1 (m1m2) variant genotypes increased the risk of oral and pharyngeal cancers (AdjOR: 1.3, 95% CI: 1.0–1.7), similar observations were made when the homozygous CYP1A1 variant (m2m2) was considered (AdjOR: 1.9, 95% CI: 1.0–3.9—Table 8). This marginal association was also observed in never smokers, but not in current or ever smokers. Similarly, the GSTM1 null plus the CYP1A1 m1m1 or m1m2 genotypes were marginally associated with the risk of oral and pharyngeal cancers in never smokers (AdjOR: 1.4, 95% CI: 0.9–2.0) but not in ever smokers (AdjOR: 1.3, 95% CI: 0.8–2.2). When oral cavity and pharyngeal cancer case subjects were examined independently, the interaction between the GSTM1 null and CYP1A1 MspI polymorphism was observed for oral cancer but not for cancer of the pharynx (Table 8).

Table 8 Overall and stratified odds ratios of the association of GSTM1/CYP1A1 MspI polymorphism with oral and pharyngeal cancers–pooled analysis
Full size table

DISCUSSION

To our knowledge, this is the first meta-analysis and pooled analysis carried out to assess the role of GSTM1 and CYP1A1 in oral and pharyngeal cancers and to evaluate potential gene-gene and gene-environment joint effects. The results obtained in this study support the hypothesis that GSTM1 deletion and certain CYP1A1 polymorphisms may play a role in the carcinogenesis process leading to oral and pharyngeal cancers. Both the meta-analysis and pooled analysis showed a significant association between oral and pharyngeal cancer and the homozygous variant genotype of the CYP1A1 MspI polymorphism. In addition, the data suggest that the combined effect of GSTM1 and CYP1A1 may be associated with oral and pharyngeal cancers. In the meta-analysis, the GSTM1 null genotype was not found to be associated with oral and pharyngeal tumors in whites. Sensitivity analysis of the Asian studies identified a data set that determined the heterogeneity. This result suggests that differences in oral and pharyngeal cancer risk factors may be present according to the geographic origin of the subjects. Ethnic differences in the association between metabolic polymorphisms and tobacco related cancers may be related to gene-gene interactions, different linkages to the polymorphisms determining oral and pharyngeal cancer risk, and different lifestyles. For example other forms of tobacco in addition to tobacco smoke, such as chewed tobacco with areca nut or wrapped in betel quid or pan101 are used in certain geographic areas. We were unable to evaluate the other ethnic groups because of heterogeneity among the studies included in this very mixed stratum.

Previous meta-analysis and pooled analysis have reported an association between the GSTM1 null genotype and head and neck tumors,40–42 but did not analyze ethnic specific or subsite specific differences. We were able to evaluate ethnic specific and subsite specific differences in the pooled analysis. We confirmed that there was no association of the GSTM1 null genotype with oral and pharyngeal cancers in whites. In contrast to the meta-analysis, there was also no association observed for Asians (OR: 1.2, 95% CI: 0.8–1.8). This difference in result may be attributed to differences in the number of subjects in the meta-analysis and pooled analysis (2313 Asian subjects in the meta-analysis versus 681 in the pooled analysis). Although not statistically significant, African American and African populations seemed to be almost two times more likely to have the GSTM1 null genotype. (Adj OR: 1.7, 95% CI: 0.9–3.3). This lack of statistical significance might also be attributed to the small number of African American and African subjects included in this pooled analysis.

Although the head and neck tumors have been historically grouped together because of the similar risk factors involved in their etiology, several authors suggest that the role of genetic susceptibility might be different in the head and neck subsites.44,83,92,94 The oral cavity, pharynx, and larynx are unique structures with different functions and possibly different sensitivities to carcinogens, especially alcohol and tobacco. Studies suggest that HPV may be the etiologic agent involved in most pharyngeal tumors (particularly those in the oropharynx).102,103 The presence of HPV along with the polymorphisms of the genes in question would certainly be relevant to our analysis. However, these data were not available in the studies included in this meta-analysis and pooled analysis. In the pooled analysis, a difference in risk for oral and pharyngeal tumors was seen for the CYP1A1 MspI variant, with oral cavity tumors statistically significantly associated with the m1m2 and m2m2 variant genotypes. We also observed that the combination of GSTM1 deletion and CYP1A1 MspI variant was significantly associated with oral cavity cancer but not with pharyngeal cancer.

There is great discrepancy in the literature as to the association of CYP1A1 genotypes with various smoking related cancers.12,20,41,43,104 The pooled analysis results confirm the association found in the meta-analysis for the variant allele of the CYP1A1 MspI polymorphism (m2/m2) and oral and pharyngeal cancers. Regarding the CYP1A1 exon 7 polymorphism, the pooled analysis revealed that the association of the Val/Val genotype with oral and pharyngeal cancers was limited only to ever smokers. One caveat is the possibility that individuals could have been misclassified because most of the earlier studies used a laboratory method that may not accurately distinguish between the exon 7 variant alleles having a C2455 base change and another recently described allele having a C2453 base change.20

The pooled analysis showed a role of tobacco consumption on the association between GSTM1 deletion and oral and pharyngeal cancer, that could be explained by the involvement of this enzyme in the metabolism of PAHs. However, there is no consistent evidence supporting this association. Some studies have found a higher level of DNA adducts and chromosome damage in lymphocytes of coke oven workers, bus drivers and tobacco smokers who lack the GSTM1 gene,24,105–108 whereas others failed to find a significant relationship.109,110 The same can be said for CYP1A1 polymorphisms.24,109,111,112 When we stratified the pooled analysis by smoking status we also observed that combined effects of GSTM1 null and CYP1A1 MspI were only present among nonsmokers. This might seem controversial because it has been demonstrated that smokers with high activating CYP1A1/low deactivating GSTM1 genotypes tend to have higher benzo[a] pyrene diolepoxide-DNA adducts.24,113,114 It has been suggested that the role of CYP1A1 and GSTM1 on lung cancer risk might be more important at low levels of exposure, but these findings need further investigation.43 Other risk factors such as alcohol must be into account. Alcohol might act as a solvent for other carcinogens, or perhaps generate and exacerbate coincident inflammation and modify the effect of susceptibility for tobacco.8,115 It might also be recommendable to assess the combined effects among other polymorphisms of the GST and CYP genes (GSTM3, GSTT1, GSTP1, CYP1A2), and of other genes involved in the detoxification of tobacco and alcohol such as N-acetyltransferases (NAT1, NAT2), microsomal epoxide hydrolase, UDP-glucuronosyltransferases, and alcohol dehydrogenase.20,41,100,116–120

The presence of heterogeneity and/or publication bias may compromise the interpretation of the meta-analyses and result in an erroneous and potentially misleading conclusion. We performed sensitivity and stratified analyses to identify the sources of heterogeneity. Potential sources of heterogeneity include ethnic group, sample size, tumor location, case-control recruitment and tobacco and alcohol consumption, most of which were easily evaluated in the pooled analysis. A general limitation to the results obtained with both the meta-analysis and the pooled analysis is the potential selection bias that may have been introduced by a poorly defined study base. Some of the publications do not provide sufficient details on the characteristics of the cases and controls, the way controls have been recruited or even the period where this occurred.66,77,79,85,88,92,94 In some hospital-based studies information on the causes for hospital admission were not provided. Nevertheless, we were able to evaluate the influence of control group source in this analysis.

There were 18 published studies that were excluded from the meta-analysis because they included laryngeal cases and did not provide site-specific data.53–69 This unavoidable exclusion was a major loss of the literature. Efforts were made to obtain these datasets for inclusion in the pooled analysis; we were successful in obtaining 6 of the 18 datasets.57,59,65,66,69,87 However, the potential for publication bias in the pooled analysis cannot be dismissed because the datasets did not entirely represent all of the published studies. Nonetheless, we did not observe any evidence of publication bias for the overall associations of GSTM1 or CYP1A1 with oral and pharyngeal cancers.

An important shortcoming to the investigation of the gene-environment effects is the possibility of misclassification of exposure. The categorization of individuals as never/ex/current/ever smokers could be inaccurate and not sufficiently standardized across studies.77,79,81,88,92,94 Misclassification of exposure could lead to biased results so this must be taken into account when interpreting the findings. It would be preferable to further characterize tobacco consumption as lifetime exposure (pack-years), but in the present meta-analysis and pooled analysis this was not possible because of the heterogeneous categorization of the smoking habits. In the majority of studies there was no information of alcohol intake, thus making it impossible to stratify for this factor.

Laboratory methods

The methods for determining the gene polymorphisms discussed in this review are described in each article. The majority of the studies used genomic DNA extracted from lymphocytes with PCR as the method for genotyping.

CONCLUSIONS

Overall, the association of GSTM1 deletion and oral and pharyngeal cancers may be dependent upon ethnicity. A possible association observed for Asians and African American/African groups and not for whites cannot be ruled out. The CYP1A1 exon 7 polymorphism was associated with oral and pharyngeal cancer only for ever smokers, when studied independently in the pooled analysis, although the CYP1A1 MspI variant homozygote allele (m2/m2) was significantly associated with this cancer in both the meta-analysis and pooled analysis. When analyzing the complete genotype of GSTM1 deletion and CYP1A1 MspI polymorphism, the risk of oral and pharyngeal cancers seems to be higher for never smokers than for ever smokers. It should be highlighted that the results of the pooled analysis varied according to the type of controls considered, indicating that a selection bias might be present in some studies and therefore the results should be considered with caution. There is no indication at this point for population testing of these genes as risk factors for oral and pharyngeal cancer.