GSTM1, GSTT1 and Risk of Squamous Cell Carcinoma of the Head and Neck
by Stacy A. Geisler^{1, 2}, Andrew F. Olshan^{1, 3}

Address correspondence to:
Stacy Geisler, DDS
Department of Epidemiology
CB#7400, School of Public Health
University of North Carolina
Chapel Hill, NC 27599-7400

March 8, 2001 (Updated August 10, 2004)

Squamous cell carcinoma of the head and neck (SCCHN) is a group of cancers of epithelial origin that may provide an ideal model for the study of gene-environment interaction. SCCHN includes squamous cell carcinomas of the oral cavity, pharynx, and larynx. Approximately 90% of the attributable risk for oral cancer and 80% of the attributable risk for larynx cancer results from tobacco use. Tobacco smoking has been demonstrated to increase risk of SCCHN in a dose-response fashion. Polymorphisms of carcinogen metabolizing enzymes (CMEs), known to be involved in metabolism of carcinogens found in tobacco smoke, are relatively common in most populations.

This paper provides a concise review of the twenty-four published studies that evaluated the risk of SCCHN in relation to two deletion polymorphisms of the glutathione S-transferase family: GSTM1 and GSTT1. Patterns of risk based on site of tumor and nationality are presented, as are some methodological weaknesses of the studies. The results of these studies are inconsistent with some reporting weak to moderate associations and others finding no elevation in risk for the main effect of the gene. Few studies have directly evaluated the interaction with tobacco. Well-designed population-based studies of adequate size are needed.

The glutathione S-transferases (GSTs) are a family of enzymes known to play an important role in the detoxification of several carcinogens found in tobacco smoke (1) . GSTs are dimeric proteins that catalyze conjugation reactions between glutathione and tobacco smoke substrates such as aromatic heterocyclic radicals and epoxides. Conjugation facilitates excretion, and thus constitutes a detoxification step. In addition to their role in phase II detoxification, GSTs also modulate the induction of other enzymes and proteins important for cellular functions such as DNA repair (1) . This class of enzymes is therefore important for maintaining cellular genomic integrity and as a result, may play an important role in cancer susceptibility.

GST enzymes are coded for at five distinct loci known as alpha, mu, theta, pi, and gamma. Two loci in particular, GSTM1 and GSTT1, may be of relevance for squamous cell carcinoma of the head and neck (SCCHN) susceptibility. The GSTM1 locus has been mapped on chromosome 1p13.3 while the GSTT1 locus exists on chromosome 22q11.2. Individuals with homozygous deletions of either the GSTM1 locus or the GSTT1 locus have no enzymatic functional activity of the respective enzyme. This has been confirmed by phenotype assays that have demonstrated 94% or greater concordance between phenotype and genotype (2-4) . Deletion variants of GSTM1 and GSTT1 that result in no functional enzymatic activity for each locus have been characterized.

Three alleles have been identified at the GSTM1 locus: one deletion allele and two others (GSTM1a and GSTM1b) that differ by C®G substitution at base position 534 (5, 6) . This C®G substitution at base position 534 results in the substitution Lys®Asn at amino acid 172 (5) . The Lys®Asn substitution results in no functional difference between the two alleles. As a result, GSTM1a and GSTM1b are categorized together as the positive conjugator phenotype. Two alleles have been identified at the GSTT1 locus; one functional and the other non functional (7) . Individuals who are of the homozygous deletion genotype are categorized as the negative conjugator phenotype while those individuals who carry either one or both functional alleles are grouped into the positive conjugator phenotype (5) .

Two observations suggest a role for GSTM1 or GSTT1 genotypes and SCCHN susceptibility. First, exposure to tobacco smoke is the most important risk factor for SCCHN (8) . Tobacco smoke is known to contain at least 55 carcinogens that can be grouped into three classes: polycyclic aromatic hydrocarbons, N-Nitrosamines, and Asz-arenes (9, 10) . Of the polycyclic aromatic hydrocarbons, benzo [a] pyrene-7, 8 dihydrodiol-9, 10 oxide (benzo [a] pyrene) is the most studied. Activation of benzo [a] pyrene results in its transformation into 7,8-diol-9, 10 epoxide (BPDE), a known substrate for the GSTM1 enzyme (11) .

Metabolism of carcinogens such as benzo [a] pyrene involves a balance of activation steps that produces reactive intermediates, and detoxification steps that produce water soluble, excretable compounds. Activation is often mediated by the cytochrome p450s pathway and can result in the formation of compounds that can bind covalently to DNA, forming what are known as adducts. Accumulation of DNA adducts at critical loci such as oncogenes or tumor suppressor genes can lead to somatic mutation and disregulation of the cell cycle (10) . Individuals who do not have the ability to produce the GSTM1 enzyme potentially accumulate more DNA adducts through their inefficiency in excreting activated carcinogens such as BPDE.

Other tobacco carcinogens such as epoxybutanes and ethylene oxide are known substrates for GSTT1 (11, 12) . GSTT1, like GSTM1, is known to play a role in phase II detoxification of carcinogens found in tobacco smoke as well as other carcinogens found in pesticides such as halomethanes and methyl bromide (1, 13) . Unlike the GSTM1 enzyme however, GSTT1 has both detoxification and activation roles (1, 14) . For example, GSTT1 is known to activate dihalomethanes to dichloromethane (DCM) which has been shown to cause liver and lung tumors in mice (1, 7) .

Unlike any other member of the GST family, GSTT1 is expressed not only in the adult liver but also in human erythrocytes, and as a result, is believed to play a more global role than GSTM1 in detoxification of carcinogens in the body (14) . This multifactorial role of the GSTT1 enzyme is believed to reflect its heritage as the ancestral progenitor gene for all mammalian GST enzymes (1) . The presence of GSTT1 enzyme within red blood cells may allow for red cells to act as a detoxification sink among those who are able to synthesize the enzyme (1) . Interestingly, if the capacity for removal of detoxification products from the circulation is exceeded among those with GSTT1 functionality, the risk of carcinogenesis may be increased as compared to those who have no function of the enzyme (1) .

The second observation suggesting that GSTM1 or GSTT1 genotypes are important for SCCHN susceptibility is that GST enzymes are expressed in the squamous mucosa of the head and neck with some site specificity (15-17) . For example, normal and malignant squamous cells of the larynx have been shown to express the GST mu isoform in the highest concentration as compared to GST alpha, pi, gamma or theta (18-21) . GSTP1 is found in the greatest concentration in the oral and pharyngeal mucosa of the head and neck as compared to the other GST enzymes (22-25) .

An extensive review of gene variants for GSTM1 and GSTT1 have been previously published and will only be briefly updated here (5, 26) . Medline and PubMed were searched using the keywords “glutathione s-transferases”, “GSTM1” and “GSTT1”. Reference lists were also reviewed from published articles. Papers written in English and published between 1993-2000 were reviewed.

The majority of the studies reviewed were case-control in design. Variation in frequencies reported among the same ethnic groups may be due to differences in study size and source of control group. Studies using hospital or other non-population controls may not represent the true genotype distribution for a given population.

In the United States, the reported range of the GSTM1 deletion genotype varies by ethnic group. Reported frequencies from hospital based case-control studies range from 23-41% for individuals of African descent, 32-53% for individuals of Asian descent, 40-53% for those of Hispanic descent, and 35-62% for those of European descent (26, 27) . Several population-based studies have reported prevalences ranging from 48%-57% for the GSTM1 deletion genotype among American Caucasians (28-31) .

South American case-control (non-population based) studies have reported frequencies of 21% for Chileans (32) and 55% for Caucasian Brazilians, 33% for Black Brazilians, and 20% for Amazonian Brazilians (33) .

European case control studies have indicated variation in the frequency of the GSTM1 deletion genotype. Among the French, 46% have been reported to carry the null genotype (34) . A large cross-sectional study conducted among Italians reported a frequency of 53% (35) ; two studies conducted in Hungary and the Slovak Republic measured frequencies of 44% and 50% (36, 37) . Finally, a population-based study conducted in Finland found a prevalence of 40% for the GSTM1 deletion genotype.

Groups such as Pacific Islanders and Malasians have a reported GSTM1 deletion genotype frequency of 62-100%. Other Asian populations such as Japanese and Chinese also have high frequency of GSTM1 deletions. Reported frequencies range from 48-50% and 35-63% respectively (5) . A population-based study conducted among Chinese reported a frequency of 51% for the GSTM1 deletion genotype (38) . Two Korean case-control studies found frequencies of 53% and 56% for the GSTM1 deletion genotype (39, 40) .

Studies of GSTT1 null genotype demonstrate that in the United States, deletion of GSTT1 is less common than the GSTM1 deletion genotype. Among those of European ancestry, 15-31% have no functional GSTT1 enzyme. African descendents have frequencies ranging from 22-29% while those of Hispanic origin carry GSTT1 deletions 10-12% (26, 27, 30, 31) .

European studies have reported that the GSTT1 deletion genotype was present among 21% of Italians and 28% of Slovakians (35, 37) . One South American study found that 19% of both Caucasian and Black Brazilians had the deletion genotype compared to 11% of Amazonian Brazilians (33) .

Asians have the highest reported GSTT1 deletion genotype. One study reported 58% of Chinese and 38% of Malaysians have the GSTT1 null genotype (41) ; two case-control studies measured 42% and 46% among Koreans (39, 40) . However, a recent population-based study conducted among Chinese found a prevalence of 46% for the GSTT1 deletion genotype among their study subjects (38).

SCCHN is a group of cancers defined by their anatomic location (oral cavity, pharynx, and larynx) and their common cell of origin (squamous cell). Roughly three times as many incident cases of oral cavity and pharynx cancer are diagnosed each year in the US as compared to incident cases of larynx cancer (42) . For the year 2000, approximately 40,400 incident cases of SCCHN will have been diagnosed in the United States and 20,400 deaths will have occurred from it (42) .

Worldwide, it has been estimated that approximately 500,000 incident cases are diagnosed each year (43) . Within the developing world, SCCHN represents the third most common cancer among men and the fourth most common among women (44) . Five-year survival has remained unchanged during the past five decades: approximately 47% of patients with oral or pharyngeal squamous cell carcinoma die within five years after diagnosis and 44% of patients with laryngeal squamous cell carcinoma die five years after diagnosis (42) .

Tobacco smoking is the strongest risk factor for SCCHN. Various population-based studies of male cigarette smokers have reported relative risks of 3-13 for ever-smokers (45) . When the amount of tobacco smoked is examined, a dose-response trend is demonstrated. Relative risks, adjusted for alcohol use, of 1.6 95% confidence interval (0.9, 2.7) for light smokers (less than 20 cigarettes per day for 20 or more years), 2.8 (1.8, 4.3) for moderate smokers (20-39 cigarettes smoked for 20 or more years), and 4.4 (2.7, 7.2) for heavy smokers (40 or more cigarettes smoked for 20 or more years) have been found (45) . In spite of a lower incidence as compared to men (roughly twice as many men are diagnosed with incident disease as compared to women in the United States), it has been suggested that women have a relatively increased risk for SCCHN per tobacco smoke dose of carcinogens (46-52) . Relative risks of 3.0 (1.9, 5.2) for light smokers, 4.4 (2.7, 7.2) for moderate smokers, and 10.2 (5.2, 20.4) for heavy smokers have been measured among women (45) .

Alcohol consumption is also linked to an increased risk of SCCHN. For those men and women who consumed greater than thirty drinks of alcohol per week, the risk of developing SCCHN was nine times that of a non-drinker (8) . Among nonsmokers, odds ratios (OR) of 1.9 (0.4, 9.6), 2.3 (0.4, 12.4), and 9.1(1.7, 48.5) have been demonstrated for light, moderate, and heavy alcohol consumers as compared to abstainers (53) .

Evidence of synergism is seen among individuals who smoke tobacco and drink alcohol. Relative risks of approximately 40 have been found among those who smoke 40 or more cigarettes a day and consume 30 drinks per week (45) . A recent case-control study conducted in Brazil among 784 cases of SCCHN measured an odds ratio of 20 for those with the greatest cumulative measures of alcohol and tobacco (53) . Blot et al. has estimated that approximately 75% of the attributable risk of SCCHN results from the combined effects of tobacco and alcohol (8) .

Human papillomavirus (HPV) may play a role in the etiology of SCCHN. Over thirty studies (most of which have been case-series) have been conducted examining the association between SCCHN and HPV genomic DNA (54) . Use of different molecular methods in identification of HPV makes comparison of these studies difficult; however three of the larger studies suggested an increase in risk for oral cancer among those infected with high risk HPV types 16 and 18 (55-57) . The overall estimates of high-risk HPV prevalence among individuals with SCCHN varies from 8-100% with an average prevalence of 35% when PCR (polymerase chain reaction) methods are used to detect the virus (58) . The exact role of HPV in SCCHN etiology remains unclear.

Epidemiological studies have demonstrated increased risk of SCCHN among the elderly (59, 60) , African Americans (61, 62) , patients of low socioeconomic status (63) , and among certain occupations (50, 64) . Diets poor in fruits and vegetables have also been identified as a risk factor for SCCHN (59, 65-68) . Use of other tobacco products such as chewing tobacco and snuff have also been identified as risk factors for oral cavity and pharynx cancer, as have use of alcohol based products such as mouthwash (49, 52, 69-71) .

The search strategy employed for identifying papers for review is the same as the strategy defined in the gene variant section. Additional search words included “head and neck cancer”. Papers published from 1993-2000 and written in English were considered eligible. Studies of squamous cell carcinoma of the esophagus were excluded.

Twenty-four hospital based case-control studies of GSTM1 and GSTT1 and risk of SCCHN have been published to date. Results of these studies are individually summarized in Tables 1-3. Two representative studies have been selected for more detailed discussion: one conducted in the United States and another conducted in Germany. Trends of results based on tumor site (oral/pharynx and larynx) are presented since GSTs can demonstrate site specificity and there is heterogeneity in risk based on tumor site. Trends based on nationality will also be summarized since frequency of GST polymorphisms varies by ethnicity, as well as the species of tobacco grown and smoked (5, 9) . A brief discussion of methodological weaknesses and how they may influence the validity of the trends will be given.

click to view

Table 1: Carcinogen Metabolizing Enzymes and Risk of Squamous Cell Carcinoma for the Americas

click to view

Table 2 : Carcinogen Metabolizing Enzymes and Risk of Squamous Cell Carcinoma for Europe

click to view

Table 3 : Carcinogen Metabolizing Enzymes and Risk of Squamous Cell Carcinoma for Asia

GSTM1 overview

The earliest published study of GSTM1 and SCCHN focused on the enzyme phenotype. A study of laryngeal cancer reported an odds ratio of 2.5 (1.8, 3.1) for persons lacking phenotypic expression of the GSTM1 enzyme (72) . Results published later by the same group reported an OR of 1.9 (1.2, 3.1) for GSTM1 null phenotype and risk of larynx cancer (73) . Coutelle et al reported an OR of 4.7 (1.0, 21.8) for larynx cancer among those who lacked GSTM1 enzyme expression and an OR of 1.8 (0.5, 6.2) for cancer of the oral cavity or pharynx after adjustment for age (74) .

Twenty-one published studies have examined the risk of SCCHN and the GSTM1 deletion genotype. Thirteen have reported odds ratios between 0.9 and 1.3 for the GSTM1 deletion polymorphism (75-87) while the remaining eight have reported odds ratios between 1.4 and 3.9. (88-95) . Among the larger studies (those with 150 cases or more), five have found odds ratios of ranging from 1.4 to 3.9 (89, 90, 92, 95, 96) while four have measured odds ratios ranging from 0.9 to 1.2 (79, 80, 85, 86) .

Two representative studies. Cheng et al published one of the largest American studies conducted to date in 1999. Cases were recruited from outpatients in the Department of Head and Neck Surgery at MD Anderson Cancer Center between May 1995 and April 1998. Details of the sampling strategy for selection of cases were not given and it was not stated whether the 162 cases recruited for the study represented incident or prevalent disease (92) .

Controls were selected from two sources: outpatients at a health maintenance organization and spouses of outpatients at MD Anderson Cancer Center. Controls were matched to cases based on tobacco exposure, age, gender, and ethnicity. Exposure data on tobacco and alcohol was collected by self-administered questionnaire. After adjustment for age, gender, ethnicity, tobacco and alcohol, an OR of 1.5 (1.0, 2.2) for GSTM1 was reported.

Another large study, conducted in Germany, was published in 1998. Matthias et al identified 398 cases of SCCHN diagnosed from 1994 to 1996 at the Department of Otorhinolaryngology, Virchow-Klinikum, Humboldt University, Berlin (79) . The authors did not give information as to whether the cases collected represented incident or prevalent disease. This case-control hospital based study collected almost all consecutively diagnosed cases of SCCHN at their institution. Data was also collected on those individuals who refused to participate in the study, and this information was used to evaluate whether refusal to participate was correlated with stage.

Controls were selected from outpatients in the same department who were undergoing surgery for hearing loss or septumplasty. Tobacco and alcohol exposures were measured by interview. However, the investigators were not able to ascertain exposure data on 50% of the controls. These investigators found an OR of 1.2 (0.8, 2.0) for oral/pharynx cancer and an OR of 1.0 (0.7, 1.5) for larynx cancer after adjustment for age and gender.

Trends based on nationality. The largest German studies have suggested minimal increase in risk (odds ratios of 1.2 for both studies) after adjustment for age and gender (79, 88) . Three out of the four largest American studies have found odds ratios of 1.4-3.1 (89, 92, 95) . Park et al reported a strong association for risk of oral cancer as modified by race. An odds ratio of 3.1 (1.1, 8.5) for African Americans and an odds ratio of 1.4 (0.7, 2.8) among Caucasians was reported after adjustment for tobacco use, alcohol consumption, and site of subject recruitment. Among Japanese studies, the largest has shown an increased risk for larynx (OR 3.9) and oral/pharyngeal cancers (OR 1.9) among smokers after adjustment for age (90) .

Trends based on site of tumor. Among those studies that have examined the risk of oral cavity squamous cell carcinoma, the majority of the studies have found no association with the GSTM1 deletion genotype (75-77, 81, 82, 84) . It is worth noting however, that among the Japanese, the majority of oral cancer studies have found an association with odds ratios ranging from 1.7 to 2.2 (90, 93, 94) . Katoh et al reported an OR of 1.7 (1.0, 2.8) for risk of oral cancer for those with the GSTM1 deletion genotype after adjustment for age and gender. Sato et al calculated an unadjusted OR of 2.2 (1.4, 3.6) for risk of oral cancer among those with the GSTM1 deletion genotype. The study by Kihara et al confirmed the results of Katoh et al with an age adjusted OR of 1.9 (0.8, 4.5) among cases of oral/pharyngeal cancer.

Three studies have found an association with laryngeal squamous cell carcinoma and the GSTM1 deletion genotype with odds ratios ranging from 1.6 to 3.9 (88, 90, 91) . Kihara et al reported that patients who smoked and carried the GSTM1 deletion genotype were almost four times more likely to be diagnosed with squamous cell carcinoma of the larynx before age 60 as compared to controls. Among 129 cases of larynx cancer and 172 controls in France, an odds ratio of 1.6 (1.0, 2.8) was detected after adjustment for age, gender, years of smoking, smoking status, daily consumption of tobacco, drinking status, and daily consumption of alcohol (91) . Finally, the largest of the three studies, conducted in Germany with 269 patients and 216 controls, found an unadjusted odds ratio of 2.8 (1.1, 6.4) (88) .

GSTT1 overview

Fourteen studies have examined the GSTT1 deletion genotype and risk of SCCHN. Six have suggested an increase in risk with odds ratios ranging from 1.4 to 2.6 (79, 82, 87, 89, 91, 92) . Other studies however have reported odds ratios of 0.5 to 1.2 (75, 76, 80, 81, 85, 86, 88, 93) .

Two representative studies. Details of the Cheng et al and Matthias et al studies were presented in the GSTM1 section. Cheng et al found an OR of 2.3 (1.4, 3.6) for those with the GSTT1 deletion genotype after adjusting for age, gender, ethnicity, smoking status, and alcohol status (92) . Matthias et al demonstrated an OR of 1.5 (0.9, 2.5) for oral/pharyngeal cancer and an OR of 0.9 (0.5, 1.4) for larynx cancer after adjustment for age and gender (79) .

Trends based on site of tumor. Of the studies that conducted a tumor site-specific analysis, two studies demonstrated an increased risk for oral squamous cell carcinoma and the GSTT1 deletion genotype (OR 2.0 and 1.5 respectively) (79, 82) while four studies did not (75, 76, 81, 93) . Among the studies that did not find an association, it is worth noting that the largest contain only 100 cases, and two of the studies contained less than fifty cases.

For cancers of the larynx, two studies have reported conflicting results. Jahnke et al reported an unadjusted odds ratio of 0.5 (0.2, 1.1) for those with deletion of the GSTT1 gene (88) while Jourenkova et al reported an odds ratio of 1.4 (0.7, 2.9) after adjustment for gender, age, duration of smoking (years), smoking status, daily consumption of tobacco in grams per day, and drinking status (91) .

Trends based on nationality. No obvious trends based on nationality have been noted for the GSTT1 deletion genotype and risk of SCCHN.

Summary

In summary, the results of the studies reviewed are inconsistent with some studies reporting weak to moderate associations and others finding no elevation in risk. Thus, the evidence for the role of GSTM1 and GSTT1 and the risk of SCCHN is inconclusive.

Methodological weaknesses of studies

A general methodological concern of the studies reviewed was the potential selection bias that may have been introduced by a poorly defined study base. Failure to properly sample from the base in a hospital case-control study can bias gene-environment interactions if the controls do not reflect the exposure and/or genotype distributions of the source population. Several of the studies reviewed employed controls that were either persons with other diseases associated with the exposure or other individuals such as friends, spouses or volunteers that may have biased exposure distributions (77, 86, 87, 89, 92) Only one study reviewed used a population-based sampling frame (76) .

Some studies employed matching of controls to cases (72, 76, 77, 82, 84, 85, 87, 89, 92, 94, 95) . Matching is often employed to increase the efficiency of the statistical adjustment of confounding factors. However, selection bias and residual confounding may be introduced when matching factors are not accounted for in the analysis (97) . Several studies reviewed did not adjust for matching factors (82, 84, 87, 95) .

Selection bias may also be introduced by the use of prevalent rather than incident cases (or a combination of prevalent and incident cases). When a mixture of incident and prevalent cases are used, differences in the genotype distribution between cases and controls might be due to the possible effects of the genotypes on survival rather than on the etiology of the disease of interest (98) . Identification of incident cases in SCCHN can be particularly challenging since patients can present with multiple primaries or a second primary after an index diagnosis. Most studies were not clear as to whether cases represented the first diagnosis of SCCHN and at least two acknowledged that a mixture of incident and prevalent cases were included (77, 87) .

All of the studies conducted to date have not been able to assess gene-environment interaction with precision due to limited statistical power. In addition to adequate sample size, assessment of gene-environment interaction also depends upon the accurate and detailed measurement of exposures, and the proper statistical evaluation of interaction on the multiplicative and additive scales.

In general, most case-control studies will require a total sample size of approximately 1000 individuals to achieve 80% power when the odds ratio for exposure effect among those without the “at-risk” genotype is 1.5 and the interaction effect is 3.0 or greater (99) . The largest studies reviewed consist of less than 400 cases and total less than 700 individuals.

Assessment of gene-environment interaction begins with satisfactory measurement of environmental exposures. Misclassification of exposure, in this case tobacco, can have important effects in gene-environment studies (100) . Several studies reviewed either neglected to report history of tobacco use (75, 80, 88, 89) or collapsed tobacco smoking into a binary variable for analysis (72, 73, 78, 83, 90, 93) . One study measured tobacco exposure as current/ former/never; thus neither measuring dose nor duration effectively (92) .

Measurement of tobacco smoking as a binary variable (such as ever/never) is rarely appropriate since a broad range of exposure levels will be grouped together using this strategy. Failing to measure both amount (dose) and length (duration) of lifetime tobacco exposures creates heterogeneity in the assessment of risk. Inaccurate categorization of tobacco exposures may ultimately prevent researchers from identifying genetically susceptibility individuals who may have increased risk to lower dose exposures. Additionally, heterogeneous categorization makes comparison across studies difficult.

Among studies that measured dose and duration of tobacco smoking, several adjusted for tobacco in the evaluation of the gene, rather than directly assessing the interaction in the analysis (76, 79, 82, 91, 93) . Adjustment for the exposure of interest in estimating the main effect of the genotype falls short of the complete assessment of gene-exposure interaction. Full description of a possible gene-exposure interaction requires an epidemiological and statistical evaluation of interaction (97) .

Modest evidence of interaction has been shown with imprecise estimates of effect for risk of SCCHN and GSTM1 null genotype among studies that have measured dose and duration of tobacco exposure (85, 94, 95) . After adjustment for age and gender, Sato et al calculated odds ratios of 3.1 (1.6, 5.9), 3.9 (1.6, 9.1), and 16.2 (4.3, 61.0) for risk of oral cancer for those with the GSTM1 deletion genotype and increasing lifetime cigarette dose (94) .

Among those studies that have evaluated gene-environment interaction for the GSTT1 deletion genotype, Olshan et al demonstrated an increasing risk of SCCHN per dose of tobacco smoking after adjustment for age, race, gender, and average number of drinks of alcohol per week. For those with the GSTT1 deletion genotype and never smokers, the risk of SCCHN was 2.7 (0.5, 12.9) as compared to never smokers without the GSTT1 deletion genotype. Less than one pack per day smokers had an OR of 3.7 (0.7, 19.4) while one pack per day or greater smokers had an OR of 7.0 (2.2, 22.0) as compared to never smokers without the GSTT1 deletion genotype. (85) .

Finally, a few studies have demonstrated increased risks for patients who have loss of function for combinations of GSTT1 and GSTM1. These studies have demonstrated that persons who are deficient in multiple enzymatic pathways have increased risk for SCCHN (76, 91, 92) . Three studies has also suggested an increase in risk for those who have a polymorphism in a phase I enzyme such as CYP1A1, and have the GSTM1 deletion genotype (77) (94) (84) .

Final considerations

Identification of groups of individuals who smoke tobacco and may have increased susceptibility for SCCHN based on their ability to metabolize tobacco smoke carcinogens is an important goal. Increased attention needs to be given to methodological considerations such as the appropriate selection of controls, use of incident rather than prevalent cases, and adequate sample size. Measurement of lifetime exposures to tobacco (measured as both dose and duration) will help to minimize heterogeneity in the assessment of gene-environment interaction. Finally, because of the carcinogenic complexity of tobacco smoke and the multi-step nature of its metabolism, consideration should be given to including multiple phase I and phase II enzymes as measures of genetic susceptibility.

Molecular methods for determining the GSTM1 and GSTT1 null genotype have been previously published and will not be reviewed here (26) . All of the studies reviewed extracted genomic DNA from blood samples except for two studies that used exfoliated oral cells only (77) (95) and one study that used both blood and oral cells (85) . Genotyping methods used in the studies reviewed were consistent with the standard techniques employed for PCR (74-76, 78, 79, 81, 83, 86-88, 90, 94) , PCR-restriction fragment length polymorphisms (84) , and multiplex PCR (77, 80, 82, 85, 89, 91-93, 95). Internal control primers were stated for all studies.

None as of yet and not indicated

• Table 1 - Carcinogen-metabolizing enzymes and risk of squamous cell carcinoma for the Americas
• Table 1a - Carcinogen-metabolizing enzymes and risk of squamous cell carcinoma for the Americas, cont.
• Table 2 - Carcinogen Metabolizing Enzymes and Risk of Squamous Cell Carcinoma for Europe
• Table 2a - Carcinogen Metabolizing Enzymes and Risk of Squamous Cell Carcinoma for Europe, cont.
• Table 3 - Carcinogen Metabolizing Enzymes and Risk of Squamous Cell Carcinoma for Asia

List of References

National Cancer Institute (last accessed 4/2007)