Glutathione S-Transferase Polymorphisms and Risk of Second Primary Malignancy after Index Squamous Cell Carcinoma of the Head and Neck Free

Approximately 80% to 90% of squamous cell carcinoma of the head and neck (SCCHN) in the United States has historically been attributed to tobacco and alcohol (1). Cigarette smokers are estimated to be at a 10-fold increased risk over never smokers of developing SCCHN, and alcohol has been suggested both to be an independent risk factor for SCCHN and to enhance the risk associated with smoking (2). However, most smokers and drinkers never develop SCCHN, suggesting that genetic susceptibility also contributes to SCCHN etiology. Surgery, radiotherapy, and chemotherapy cure many SCCHN, but a significant cause of posttreatment morbidity and mortality is the development of second primary malignancies (SPM), estimated to occur in ∼15% of SCCHN patients (3).

Enzymes in carcinogen detoxification, nucleotide excision repair, cell cycle control, and apoptosis pathways have all been suggested to interact with tobacco-associated carcinogens to modulate interindividual susceptibility to SCCHN (4, 5). Glutathione S-transferases (GST) are a family of seven phase II metabolizing enzymes that play a central role in catalyzing the conjugation of electrophilic compounds such as environmental carcinogens to glutathione, converting them to hydrophilic compounds that are less reactive and more easily excreted (6). The frequency of homozygous deletion carriers for GSTM1 and GSTT1 ranges from 27% to 53% and 20% to 47%, respectively (6), and individuals with homozygous deletions of the GSTM1 and GSTT1 genes have no respective enzyme activity (7). Polymorphisms in the GSTP1 gene have been shown at exons 5 (Ile¹⁰⁵Val) and 6 (Ala¹¹⁴Val; ref. 8), with valine allele frequencies ranging from 18% to 42% and 5% to 9%, respectively (9). The Ile¹⁰⁵Val and Ala¹¹⁴Val polymorphisms have been shown to have both decreased affinity for electrophilic substrates in vitro (10, 11) and decreased enzymatic activity in vivo in human lung tissue (9). These polymorphisms and associated differences in enzyme activity provide a potential mechanism for increased susceptibility to smoking-related cancers, including SCCHN. We have recently reported that GST polymorphisms and specific haplotypes may modulate in vitro benzo[a]pyrene diol epoxide–induced adducts (12). Additionally, meta-analyses have suggested an increasing risk of head and neck cancer with inheritance of increasing numbers of modest risk polymorphisms of GSTM1, GSTT1, and GSTP1 (7).

Understanding susceptibility to tobacco-related malignancies holds great promise for primary cancer prevention. Additionally, identifying markers of risk for SPM among cancer survivors would greatly enhance secondary prevention, which is currently limited to rather simplistic clinical posttreatment screenings. Using a prospectively recruited cohort of 1,600 incident SCCHN patients without prior treatment or prior other malignancies, we assessed the association of genetic polymorphisms of GSTM1, GSTT1, and GSTP1 and the risk of SPM development. We hypothesize that genetic polymorphisms in these three phase II conjugation genes will individually, and more likely collectively, confer an increased risk of SPM.

Between May 1995 and December 2006, 1,600 newly diagnosed and previously untreated patients with histopathologically confirmed SCCHN were consecutively recruited as part of an ongoing prospective molecular epidemiologic study at our institution, which has been previously described (13, 14). All subjects completed Institutional Review Board–approved informed consent and were recruited without discrimination about age, sex, ethnicity, or cancer stage, except that patients with known distant metastases were excluded. Patients with any cancer history excepting nonmelanoma skin cancer were not recruited. All patients with primary sinonasal tumors, salivary gland tumors, cervical metastases of unknown origin, or tumors outside the upper aerodigestive tract were also excluded. Additionally, patients who underwent only palliative treatment were excluded. Approximately 95% of contacted patients consented to enrollment in the study. Two hundred twenty-four (14.0%) patients without either treatment or follow-up at our institution were excluded from this study cohort. Of the remaining 1,376 patients, blood samples for genotyping data were not available for 165 (12.0%) patients, and these patients were excluded for final genotype analysis.

Patients were monitored through their treatment and posttreatment course with regularly scheduled clinical and radiographic examinations. SPMs were distinguished from local recurrences based on modified criteria of Warren and Gates (15). Second lesions with different histopathologic type, or occurring >5 y following treatment for the primary tumor, or clearly separated by normal epithelium based on clinical and radiographic assessment were considered SPM. If there was discrepancy or differing opinion about the origin of the tumor, the second lesion was classified as a local recurrence rather than a SPM. Pulmonary lesions were considered SPM if they had a nonsquamous histology or if they were isolated squamous lesions >5 y from initial SCCHN and felt to be SPM by the thoracic oncologist and thoracic surgeon. SPMs were then classified as head and neck (squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, or larynx), other tobacco-associated (esophagus, lung, or bladder), or other non–tobacco-associated malignancies.

Clinical data were obtained at initial presentation and through follow-up examinations and included overall stage at presentation of the index tumor, site of the index tumor, and treatment. All patients completed at presentation an epidemiologic questionnaire, including data on alcohol and smoking status. Alcohol status was categorized as “ever drinkers” (those who had drunk at least one alcoholic beverage/week for at least 1 y during their lifetime) and “never drinkers” (those who never had such a pattern of drinking). “Ever smokers” were those who had smoked at least 100 cigarettes in their lifetime, and “never smokers” were those who had smoked <100 cigarettes in their lifetime.

The genotypes of GSTM1 null, GSTT1 null, GSTP1 Ile¹⁰⁵Val, and GSTP1 Ala¹¹⁴Val were analyzed by PCR and restriction enzyme digestion using previously published primers and methods (9, 12, 16). There was 100% concordance when 10% of the genotyping assays were repeated.

The primary end point of the study was SPM occurrence. Time to event was calculated from the date of diagnosis of the index SCCHN to the date of SPM occurrence. Patients who were not known to have an event at the date of last contact, or who were lost to follow-up, or who died following the initial date of diagnosis of their first primary head and neck cancer were censored. SPM-free survival was calculated as the number of months from the date of diagnosis of the index SCCHN to either SPM development or the end of the study. Kaplan-Meier curves were used to estimate SPM-free survival, and the log-rank statistic was used to evaluate significant differences (α = 0.05) in SPM-free survival between the different genotype groups. The associations between individual epidemiologic risk factors, clinical characteristics including tumor site, staging, and treatment variables, and time to SPM occurrence were initially assessed using univariate Cox proportional hazards regression models.

Epidemiologic variables in the univariate analysis, assessed at the time of diagnosis, included age in years, ethnicity, sex, and smoking and alcohol status. Clinical characteristics included tumor site, tumor stage, and treatment. The first step in building a multivariable model for time to SPM occurrence did not incorporate any interaction terms. A multivariable proportional hazards model was built using the variables that had SPM predictive potential suggested by the univariate analysis (P < 0.25). Due to epidemiologic and clinical considerations in building the model, age, sex, and ethnicity were always retained in the main-effects and final multivariable model. In building multivariable models, a stepwise search strategy was used. A threshold level of 0.25 for the likelihood ratio test was used as a cutoff to determine whether a variable could be entered into, or removed from, the regression model. Associations were quantified using hazard ratios (HR) and their 95% confidence intervals (95% CI) for developing SPM. The final fully adjusted Cox regression models included age, sex, ethnicity, and smoking and alcohol status. χ² Test was used to evaluate for differences in age, sex, ethnicity, smoking and alcohol status, cancer site, stage, and treatment between the two groups of participants who developed SPM and those who did not. For all analyses, statistical significance was set at P < 0.05, and all tests were two sided. Statistical Analysis System version 9.1 was used to do all statistical analyses.

Based on previously mentioned exclusion criteria, a cohort of 1,376 patients with index SCCHN was followed prospectively for development of SPM. Median follow-up time was 26 months (range, 0-142.4 months). One hundred ten (8.0%) patients developed SPM. Of patients with SPM, 43 (39%) had second SCCHN, 38 (35%) developed other tobacco-associated cancers, and 29 (26%) developed other non–tobacco-associated malignancies. Of the 43 patients with SPM of the head and neck, 24 (56%) were synchronous SCCHN primaries. Of these 24 patients with synchronous SCCHN, 2 had bilateral oral cavity cancers, 3 had bilateral oropharyngeal cancers, 1 had bilateral hypopharyngeal cancers, and the remainder had simultaneous cancers of more than one head and neck subsite.

Demographics, exposure, and clinical variables for the total cohort of patients, the patients who did not develop SPM, and the patients who developed SPM are summarized in Table 1. The mean age at diagnosis for the total patients was 57.3 years (range, 18-94 years; median, 57 years), and the mean age of patients who developed SPM was significantly older compared with the mean age of patients who did not develop a second primary tumor (60.8 years versus 57 years, respectively; P < 0.01). Although the participants in this study were predominantly male (76.0%) and non-Hispanic Whites (84.0%), gender and ethnicity were not associated with SPM development (P = 0.734 for gender and P = 0.100 for ethnicity, respectively). Compared with the SPM-free group, patients who developed SPM were older (P < 0.001) and were more likely smokers (P = 0.022) and drinkers (P = 0.05). However, compared with the SPM-free group, patients who developed SPM had similar characteristics with respect to index cancer site (P = 0.118), index cancer stage (P = 0.830), and treatment (P = 0.982).

Genetic data were available for 1,211 (88.0%) patients in the cohort. Consistent with previously published studies (6, 9), 50.4% of patients were GSTM1 null, 20.7% were GSTT1 null, and 9.9% were both GSTM1 and GSTT1 null. Valine allele frequencies were 36.3% and 8.7% for the GSTP1 105 and 114 codons, respectively.

The association between these GST polymorphisms and risk of SPM development is shown in Table 2. Patients with the GSTM1 and GSTT1 null genotypes had no significantly higher risk of SPM development than those with the wild-type genotype. Similarly, when the GSTM1 and GSTT1 genotypes were combined, the risk of SPM development among patients with both either M1 or T1 null genotype and both M1 and T1 null genotypes was not significantly different from that among patients with GSTM1 and GSTT1 wild-type genotypes. However, when data were adjusted for age, sex, ethnicity, smoking, and alcohol (Table 2), patients with the GSTP1 105 variant genotype had a 1.7-fold elevated risk for developing SPM compared with patients with the wild-type genotype (HR, 1.7; 95% CI, 1.1-2.5). We did not observe the similar association between the patients with GSTP1 114 variant genotypes and those with GSTP1 114 wild-type genotype. After we combined the risk genotypes of GSTP1 105 and GSTP1 114 polymorphisms, the patients with 1 to 2 or 3 to 4 risk genotypes had a 1.6- or 2-fold increased risk for SPM development compared with the patients with 0 risk genotypes (HR, 1.6; 95% CI, 1.1-2.5 and HR, 2.0; 95% CI, 0.9-4.3, respectively). Moreover, the risk was significantly increased in a dose-response manner, suggesting that the two polymorphisms may have a joint effect on the risk of SPM development (P_trend = 0.018).

Results were similar when SPM risk was stratified according to SPM type (Table 3). Patients with the GSTP1 105 variant genotype had a 2.2-fold increased risk for SCCHN SPM (HR, 2.2; 95% CI, 1.1-4.6) and a 1.7-fold elevated risk for SCCHN or other tobacco-associated SPM (HR, 1.7; 95% CI, 1.0-2.7). Patients with 3 to 4 GSTP1 risk genotypes had a 3.3-fold increased risk for SCCHN SPM (HR, 3.3; 95% CI, 1.0-10.5) and a 2.3-fold elevated risk for SCCHN or other tobacco-associated SPM (HR, 2.3; 95% CI, 1.0-5.3) compared with patients with no risk genotypes. When SPM risk was stratified according to smoking status at the time of diagnosis of index SCCHN (never versus ever smokers), there was no significant difference in the HRs among these two subgroups (data not shown).

To evaluate combined effects of a panel of GST polymorphisms that act in the same carcinogen metabolizing pathway, which may amplify the effects of associations of these polymorphisms with the risk of SPM development among SCCHN patients, we combined the risk genotypes of these four polymorphisms for further combined analysis for the entire cohort patients (Table 4). We found that the distribution of the combined risk genotypes was borderline significantly or significantly different between patients who developed SPM and those who did not (P = 0.096 for quartile categories and P = 0.035 for dichotomized categories). Patients with 2, 3, or 4 risk genotypes experienced a shorter SPM-free survival compared with patients with 0 or 1 risk genotypes (log rank, P = 0.039; Fig. 1). Similarly, patients with 3 or 4 risk genotypes had a short SPM-free survival compared with patients with 0 or 1 or 2 risk genotypes (log rank, P = 0.015; Fig. 2). After adjusting for age, sex, ethnicity, smoking, and alcohol, there was a stepwise increase in SPM risk with increasing number of risk genotypes (P_trend = 0.004). Patients with 4 risk genotypes had a 2.6-fold significantly elevated risk of SPM compared with patients with 0 or 1 risk genotypes (HR, 2.6; 95% CI, 1.1-6.1). We also found that patients with 3 or 4 risk genotypes had a 1.7-fold significantly elevated risk for SPM compared with patients with 0 or 1 or 2 risk genotypes (HR, 1.7; 95% CI, 1.2-2.5).

Exposure to environmental carcinogens and genetic variation in response to this exposure have been linked to both SCCHN and SPM (7, 16, 17). As phase II detoxifying enzymes, the GSTs have been targeted as possible genetic biomarkers to predict malignancy risk. GSTs have been most consistently reported to confer increased risk for lung and bladder malignancies (6). Whereas individual studies have reported varying associations between GSTs and SCCHN, meta-analyses and pooled analyses of GSTM1, GSTT1, and GSTP1 genotypes have revealed modest associations of the GSTM1 null, GSTT1 null, and GSTP1 Ile¹⁰⁵Val polymorphism with risk of SCCHN. Furthermore, combination of these three polymorphisms into a multigenic model conferred a 2.1-fold increased risk of SCCHN (7).

The same genetic susceptibility that confers risk for index cancer may also increase risk for development of SPM, which remains a cause of significant morbidity and mortality among SCCHN patients. Small studies have previously examined the association between GSTs and SPM after index SCCHN. In a nested case-control study of 20 patients, Bongers and colleagues (18) found that expression of GST in healthy tissue in direct vicinity of an index tumor predicted SPM development. Matthias and others (19) found a significantly higher rate of the GSTT1 null genotype in 39 patients with multiple SCCHN compared with patients with a single SCCHN but did not find significant differences associated with the GSTM1 null genotype or the GSTP1 Ile¹⁰⁵Val polymorphism. Minard et al. (17) reported 50 SPMs in a cohort of 303 patients with early-stage disease, with an increased risk of SPM associated with the GSTM1 null genotype but not the GSTT1 null genotype.

GST polymorphisms have been most commonly studied with other tobacco-associated cancer sites, such as the lung and bladder. In lung cancer studies, there is growing evidence that combinations of GST polymorphisms confer increased risk of cancer (20). In a study of glutathione conjugation and DNA adduct formation, Wenzlaff et al. (21) showed that multiple differential GST isoenzymes (including GSTM1 and GSTP1) can effectively detoxify the most carcinogenic polycyclic aromatic hydrocarbons. Bolt and Their (22) found a 7-fold increased risk of small cell lung carcinoma in patients with both the GSTM1 null and the GSTP1 Val¹⁰⁵Val genotypes. Similarly, Sundberg et al. (23) found a 4-fold increased risk of lung cancer in nonsmokers carrying both the GSTM1 null and GSTP1 valine alleles. Finally, Ryberg et al. (24) noted significantly increased frequency of GSTM1 null and GSTP1 Ile¹⁰⁵Val polymorphisms among lung cancer patients when compared with controls, as well as significantly higher levels of DNA adducts in lung cancer patients carrying these two combined variant genotypes.

With 1,376 patients, 110 (8.0%) of whom developed SPM, the current investigation represents the largest cohort study to evaluate the influence of individual and combined GST polymorphisms on risk of SPM after index SCCHN. The GSTP1 Ile¹⁰⁵Val polymorphism had the strongest association with SPM development, with a 1.7-fold increased risk of SPM and a 2.2-fold increased risk of SCCHN SPM. A stepwise increase in SPM risk was observed with increasing numbers of risk genotypes of these four polymorphisms (P_trend = 0.004). After combining all four polymorphisms, patients with 4 risk genotypes of the four polymorphisms had a 2.6-fold significantly increased risk of SPM compared with patients with 0 or 1 risk genotypes. Our findings were similar in subgroup analyses limited to only ever smokers (data not shown) and when outcomes were stratified according to SPM type.

This work has several inherent limitations. This cohort includes multiple ethnicities, which have been shown to have different frequencies of GST polymorphisms (6). However, 85% of patients were non-Hispanic Whites, and ethnicity was included in the multivariate model. Secondly, only one of four polymorphisms in this investigation (GSTP1 Ile¹⁰⁵Val) had a statistically significant association with SPM risk, a finding that could be attributed to chance. However, several studies have documented the biological basis of decreased enzymatic activity and affinity for electrophilic substrates associated with the GSTP1 Ile¹⁰⁵Val polymorphism (9–11). In addition, there were trends toward higher SPM rate with both the GSTM1 null and the GSTT1 null genotypes, which could potentially have reached statistical significance with increased time of follow-up. Thirdly, whereas demographics, exposure, and clinical data for the cohort were collected prospectively, clinical outcomes including SPM were collected retrospectively without a strictly defined screening or follow-up regimen. Furthermore, follow-up time in this study may have been limited by patients with stage III and IV index cancer disease who were lost to follow-up. Finally, SPM rate (8%) was lower than expected, and the SPM sample size limits statistical power in the subgroup analyses. We suspect that the low SPM rate is due to our high prevalence of never smokers (28%) and patients with stage III and IV index cancer disease (75%), as well as our strict criteria in defining SPM.

Another limitation of the current investigation is the absence of human papillomavirus status, which could potentially influence the development of second primary cancers in patients with index SCCHN. However, smoking status is a surrogate marker for human papillomavirus status, and we have included smoking status in the multivariable model. Additionally, we found no difference in HRs in a subgroup analysis of GST polymorphisms and SPM risk among ever and never smokers. We will focus on the role of human papillomavirus in outcome of SCCHN patients for our future studies once the tumor human papillomavirus data are available.

Despite these limitations, the current investigation supports a significant role for GST in the multigenic cancer model. SCCHN and SPM result from complex interactions of multiple genetic and environmental factors over time that cannot be explained completely by a single environmental exposure or allelic variability at a single locus. As in this investigation, the strongest associations between GST and cancer susceptibility have been found when studying combined analyses of multiple GST genotypes, xenobiotics, or other genetic susceptibility genes such as those involved in phase II enzymes, DNA repair, cell cycle control, or apoptosis (25). Increased understanding of the cumulative effects of multiple genetic variants in the GST and other pathways through large multi-institutional studies such as the International Head and Neck Cancer Epidemiology Consortium will potentially help identify biomarkers of susceptibility for both index and second primary head and neck malignancies.

No potential conflicts of interest were disclosed.