TY - JOUR
AU - Kelsey, Karl T.
AB - Abstract Background: Cigarette smoking is associated with a twofold increased risk of pancreatic cancer. We conducted a population-based case–control study in six San Francisco Bay area counties from 1994 to 2001 to investigate associations between polymorphisms in genes for two carcinogen-metabolizing enzymes (cytochrome P450 1A1 [CYP1A1] and glutathione S-transferase [GST]), smoking, and adenocarcinoma of the exocrine pancreas. Methods: We used polymerase chain reaction–based methods to analyze blood samples obtained from 309 case subjects and 964 control subjects to determine their genotypes for three CYP1A1 polymorphisms (m1, m2, and m4) and for homozygous deletions of two GST genes, GSTM1 and GSTT1. Control subjects were frequency matched to case subjects by age and sex. All statistical tests were two-sided. Results: None of the genetic polymorphisms themselves affected the risk of pancreatic cancer among Caucasian study participants. However, we observed an interaction between GSTT1-null genotype and cigarette smoking among Caucasians that was more prominent among women than among men. Relative to never smokers with the GSTT1-present genotype, the age-adjusted odds ratios (ORs) of pancreatic cancer for heavy smokers with the GSTT1-null genotype were 5.0 (95% confidence interval [CI] = 1.8 to 14.5) for women and 3.2 (95% CI = 1.3 to 8.1) for men; for heavy smokers with the GSTT1-present genotype they were 2.0 (95% CI = 1.0 to 4.0) for women and 2.1 (95% CI = 1.1 to 3.9) for men. ORs for pancreatic cancer among heavy smokers with both GSTT1-null and GSTM1-null genotypes were similar in magnitude to those among heavy smokers with the GSTT1-null genotype alone. There was no evidence of an interaction between CYP1A1 polymorphisms and smoking. Conclusions: The combination of heavy smoking and a deletion polymorphism in GSTT1 is associated with an increased risk of pancreatic cancer among Caucasians, with the associations possibly stronger in women than in men. J Natl Cancer Inst 2002;94:297–306] Pancreatic cancer is the fifth leading cause of cancer deaths among men and women in the United States, but relatively little is known about its environmental etiology (1,2). Cigarette smoking has been modestly associated with risk of pancreatic cancer and is estimated to account for between 25% and 29% of pancreatic cancer incidence (3–5). One study (4) observed a higher estimate of population attributable fraction due to smoking among women (29%) than among men (26%). Experimental and epidemiologic evidence suggests that both carcinogen-induced DNA damage and cellular damage induced by inflammation play a role in pancreatic carcinogenesis (6–9). For example, aromatic amines and nitrosamines present in cigarette smoke are thought to play a role in pancreatic carcinogenesis via activation by cytochrome P-450s (6). Cigarette smoke, which has been shown to induce DNA damage and mutations in target cells, also contains and generates free radicals and oxidants (10–14). Oxidative stress and free radical generation occur in pancreatitis, an inflammatory disease of the pancreas that is itself a risk factor for pancreatic cancer and for which smoking is a risk factor (8,15–17). These observations led us to hypothesize that polymorphisms in genes that encode carcinogen-metabolizing enzymes previously shown to alter risks from either the direct action of tobacco carcinogens or from the proinflammatory oxidative effect of tobacco constituents also affect the risk of smoking-related pancreatic cancer. The human cytochrome P-450 1A1 (CYP1A1) gene, which is located on chromosome 15 at q22–q24, encodes arylhydrocarbon hydroxylase (AHH), a phase I enzyme involved in the activation of tobacco-related procarcinogens, such as polycyclic aromatic hydrocarbons (PAHs), nitrosamines, and aromatic amines. There are four previously studied polymorphisms in the human CYP1A1 gene. The CYP1A1 m1 polymorphism consists of a T-to-C substitution in the 3′ noncoding region of the CYP1A1 gene that creates an MspI restriction enzyme cleavage site. The CYP1A1 m2 polymorphism, an A-to-G substitution at nucleotide 4889 in exon 7, a region that encodes a heme-binding domain of CYP1A1, results in the Ile462Val polymorphism (18). The CYP1A1 m3 polymorphism, which was not evaluated in our study, is found only in African Americans (19). The CYP1A1 m4 polymorphism, a C-to-A substitution that results in the Thr461Asn amino acid substitution, is only 2 base pairs away from the m2 polymorphism and consequently has been less well studied phenotypically (20). Some studies (21,22) have found that CYP1A1 m2, but not CYP1A1 m1, is associated with elevated levels of inducible CYP1A1 enzyme activity. Molecular epidemiologic studies of CYP1A1 variants (23) have linked the m1 and the m2 alleles to smoking-related cancers of the lung, head and neck, and esophagus in Asian populations. In studies of tissue-specific expression of cytochrome P-450 enzymes (24), levels were greater in both pancreas and liver samples from patients with pancreatitis and pancreatic cancer than in tissue samples from apparently nondiseased organ donors. The glutathione S-transferases (GSTs) are a family of phase II isoenzymes believed to protect cells from reactive chemical intermediates and oxidative stress resulting from a wide range of electrophilic xenobiotics (e.g., tobacco-related carcinogens) and endogenous intermediates (e.g., reactive oxygen species) (25). GST expression varies between individuals, and expression is tissue and sex specific (26–32). Inheritance of null (gene deletion) alleles in the GSTM1 (chromosome 1p13.3) and GSTT1 (chromosome 22q11.2) genes is common in the population, varies by ethnicity (25,33,34), and is associated with the loss of enzyme activity and cytogenetic damage (35–37). Individuals that have either the m1 or m2 alleles of CYP1A1 and the GSTM1-null allele have higher levels of PAH and benzo[a]pyrene diolepoxide (BPDE)-DNA adducts in their leukocytes and lung tissue (38–40). In contrast, although the GSTT1 wild-type enzyme detoxifies smaller reactive hydrocarbon intermediates, such as ethylene oxide, the wild-type allele does not appear to be associated with the presence of PAH or DNA adducts (25,38). Although case–control studies (41–47) have linked homozygous gene deletions of GSTM1 and GSTT1 to susceptibility to various cancers, including lung, bladder, head and neck, colon, and basal cell carcinoma, the results have been inconsistent. We conducted a large population-based molecular epidemiologic study to investigate associations between pancreatic cancer and polymorphisms in carcinogen-metabolizing genes and other risk factors for this disease. We specifically investigated whether common genetic variants in the CYP1A1 gene, or loss-of-function deletion polymorphisms in GSTM1 or GSTT1, were associated with an altered risk of pancreatic cancer and whether any of these polymorphisms modified the effect of cigarette smoking on risk of pancreatic cancer. Subjects and Methods Study Population Case subjects had primary adenocarcinoma of the exocrine pancreas that was diagnosed between 1994 and 2001 and were identified by the Northern California Cancer Center (Union City, CA) using rapid case ascertainment. Eligible participants were 21–85 years of age, resided in one of six San Francisco Bay area counties (Alameda, Contra Costa, Marin, San Francisco, San Mateo, or Santa Clara) at the time of diagnosis, were alive at the time of ascertainment, and were able to communicate in English. Additional case subjects who met all study criteria but did not reside in one of the six San Francisco Bay area counties at the time of diagnosis (i.e., out-of-area subjects) were obtained through clinical files at the University of California, San Francisco (UCSF), Medical Center. Control subjects were identified by using the random digit dialing (RDD) telephone recruiting method and, for those who were 65 years old and older, the Health Care Finance Administration (HCFA) lists. Control subjects were frequency matched with case participants by sex and by age within 5-year categories. Out-of-area control subjects, also identified by RDD, were frequency matched to out-of-area case subjects by sex, age, and their current home telephone area code and prefix. Detailed interviews were conducted in person at the participants' homes or at a location of their choice for San Francisco Bay area participants and by telephone for out-of-area participants. Study procedures were approved by the UCSF institutional review board, and written informed consent was obtained from each of the study participants before he or she was interviewed or provided a blood sample. We interviewed 530 eligible pancreatic cancer patients for this study, which represented 65% of the 717 eligible San Francisco Bay area rapid ascertainment case subjects and 80% of the 81 eligible out-of-area case subjects. The analyses presented here are based on 309 interviewed case subjects who subsequently had their blood drawn to participate in the laboratory portion of the study and whose specimens were available for genetic testing. Blood or DNA specimens were not obtained from the remaining interviewed case subjects for the following reasons: no blood was drawn from out-of-area cases, patient was too sick, patient had died, patient refused, physician refused, or blood draw was insufficient or unsuccessful. We interviewed 1701 eligible control participants for this study. Of those, 59% were obtained by RDD recruitment within the six-county San Francisco Bay area, 4% were obtained by RDD recruitment outside the six-county San Francisco Bay area, and 37% were recruited from HCFA lists. The eligible control subjects who completed interviews represented 60% of the 1680 eligible San Francisco Bay Area RDD control participants, 69% of the 94 eligible out-of-area RDD control participants, and 53% of the 1191 eligible HCFA control participants. Analyses presented here are based on 964 control participants who subsequently gave blood as part of the laboratory portion of the study and whose specimens were available for genetic testing. Blood or DNA specimens were not obtained from the remaining interviewed control participants for the following reasons: no blood was drawn from out-of-area control participants, participant refusal, participant had moved since the interview, participant was ill, blood draw was insufficient or unsuccessful, or participant had died. The interview for all subjects included questions on tobacco use, alcohol consumption, diet, occupational exposures, family history, medical history, and demographic information. Race was self-reported according to three broadly defined categories: Caucasian, African American, and Asian. Hispanic participants were classified as Caucasian, Asian, or African American, depending on which of these racial categories was selected by the respondent. No proxy interviews were conducted. Genotyping Assays Genomic DNA was extracted from peripheral blood lymphocytes using the QIAmp DNA Blood Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. All polymerase chain reaction (PCR) assays contained 0.1 μg of genomic DNA, Taq polymerase (Applied Biosystems, Foster City, CA) and 1.5 mM (for CYP1A1 amplifications) or 2.0 mM (for GST amplifications) Mg2+ in standard PCR buffer 1 (Applied Biosystems). The CYP1A1 m1 allele was detected using a nested PCR reaction. The first PCR reaction contained 0.75 U of Taq polymerase and oligonucleotide primers OZ-1 (forward; 5′-TCACTCGTCTAAATACTCACCCTG-3′) and ZF-2 (reverse; 5′-TAGGAGTCTTGTCTCATGCCT-3′) and 12 amplification cycles of 94 °C for 30 seconds, 65 °C for 30 seconds, and 72 °C for 1 minute. The second PCR reaction contained 1.25 U of Taq polymerase and oligonucleotide primers ZF-1 (forward; 5′-CAGTGAAGAGGTGTAGCCGCT-3′) and OZ-2 (reverse; 5′-GAGGCAGGTGGATCACTTGAGCTC-3′) and 32 amplification cycles of 94 °C for 30 seconds, 65 °C for 30 seconds, and 72 °C for 1 minute. Fifteen microliters of the final PCR products were incubated with 5 U of the restriction enzyme MspI at 37 °C overnight and resolved on agarose gels. Cleaved PCR products indicated the presence of the variant CYP1A1 m1 allele. CYP1A1 m2 and m4 alleles were detected using a PCR–restriction fragment length polymorphism (RFLP) test as described by Cascorbi et al. (20). We determined the genotypes (wild-type [WT] or heterozygous deletion versus homozygous deletion) of participants at their GSTM1 and GSTT1 loci using a multiplex PCR assay and the oligonucleotide primer pairs GSTT1 (forward; 5′-TTCCTTACTGGTCCTCACATCTC-3′) and GSTT2 (reverse; 5′-TCACCGGATCATGGCCAGCA-3′), and GSTM1 (forward; 5′-CTGCCCTACTTGATTGATGGG-3′) and GSTM2 (reverse; 5′-TGCATTGTAGCAGATCATGC-3′). Each reaction also included oligonucleotide primers 1A1-A (forward; 5′-GAACTGCCACTTCAGCTGTCT-3′) and 1A1-B (reverse; 5′-CAGCTGCATTTGGAAGTGCTC-3′), to serve as an internal control for amplification of CYP1A1 gene sequences. The following PCR conditions were used (MgCl2, 2 mM): 40 cycles of 94 °C for 30 seconds, 60 °C for 30 seconds, and 72 °C for 1 minute. PCR-amplified DNA was resolved on 2.5% agarose gels containing ethidium bromide. Polymorphism Designation We used the following nomenclature to specify genotypes at specific carcinogen-metabolizing gene loci: CYP1A1 m1 (MspI, T→C): wild-type (WT)/WT, WT/m1, m1/m1; CYP1A1 m2 (exon 7, Ile→Val, A→G): Ile/Ile, Ile/Val, Val/Val; CYP1A1 m4 (Thr→Asn, C→A): Thr/Thr, Thr/Asn, Asn/Asn; GSTM1: present (WT or heterozygous deletion), null (homozygous deletion); and GSTT1: present (WT or heterozygous deletion), null (homozygous deletion). Statistical Methods Tests for Hardy-Weinberg equilibrium were conducted by comparing observed genotype frequencies with expected genotype frequencies among control subjects using a χ2 test with 1 df. Expected genotype frequencies were calculated from allele frequencies. ORs and 95% CIs were estimated using unconditional logistic regression analysis in SAS (v.8; SAS Institute, Cary, NC). All statistical tests were two-sided. For these analyses, we used data obtained at interview using structured questionnaires. Age at interview was treated as a continuous variable. Participants were classified as never smokers if they reported that they had smoked 100 or fewer cigarettes in their lifetime and had smoked pipes or cigars less than once per month for at least 6 months. Cut-points for the cigarette smoking variables, duration of smoking (i.e., 1–13, 14–26, 27–39, and ≥40 years) and pack-years (the number of packs per day multiplied by the number of years of smoking (i.e., <6.3, 6.3–20.21, 20.22–40.99, or ≥41), were based on quartiles of the distribution of each variable among control participants who were classified as smokers. We also evaluated recent smoking by restricting pack-years or duration of smoking to that within 15 calendar years of either the date of diagnosis (for case subjects) or study interview (for control subjects). ORs for cigarette smokers are presented relative to never smokers. Participants were classified as never drinkers if they never consumed one or more alcoholic drinks per month during their lifetime. Variables for any lifetime consumption of alcohol (at least one drink per month) were analyzed separately for each type of alcoholic beverage (beer, wine, or liquor) and in combination (never drinker, beer or wine, liquor only or liquor plus beer or wine, and liquor plus beer plus wine). The number of alcoholic beverages consumed per day during the previous year was analyzed for each type of drink separately (light beer, beer, white wine, red wine, or liquor) and in combination for any type of alcoholic beverage. No differences in the results were seen using these various definitions of alcohol consumption as a potential confounder in the analyses. Potential confounders were included in the multivariable models if their inclusion caused β parameter estimates to change by more than 10%. Final multivariable logistic models for smoking, metabolic polymorphisms, and pancreatic cancer included only age at interview and sex, the two variables used in matching case and control subjects. Potential confounders that were evaluated and did not change parameter estimates by more than 10% were alcohol or coffee consumption, educational level, annual household income, first-degree family history of pancreatic cancer, and a personal history of diabetes mellitus, gallbladder disease, ulcer, allergy, or vitamin B12 deficiency. Self-reported history of pancreatitis (25 case subjects, 12 control subjects) was evaluated as a possible confounder but was omitted from multivariable models presented in the tables because this condition may be an intermediate in the causal pathway between smoking and incidence of pancreatic cancer (8,16,17). However, for comparison, all final models were run with a variable for history of pancreatitis, and the pattern of results was similar to that obtained from models lacking this variable. Gene–environment, gene–gene, and gene–gene environment interactions were assessed by using stratified analyses and by evaluating departures from additive effects. Because both approaches gave the same results, we present only departures from additive effects. We evaluated departures from additive effects of two variables by coding a new variable with a common referent group based on a priori hypotheses. For example, to evaluate the combined effect of the GSTM1-null genotype and pack-years of smoking, we coded a new variable with the following six categories: GSTM1-present/never smoker (referent group), GSTM1-null/never smoker, GSTM1-present/smoker with less than 41 pack-years, GSTM1-null/smoker with less than 41 pack-years, GSTM1-present/smoker with greater than or equal to 41 pack-years, and GSTM1-null/smoker with greater than or equal to 41 pack-years. ORs for the combined effect of genotype and cigarette smoking were estimated using unconditional multivariable logistic models (PROC LOGISTIC) in SAS. The magnitude of an interaction effect was determined by estimating the age- and sex-adjusted interaction contrast ratio (ICR) and 95% CIs using PROC LOGISTIC in SAS (48). We calculated the ICR using the following formula:  \[\mathrm{ICR\ =\ RR_{11}\ {-}\ RR_{10}\ {-}\ RR_{01}\ +\ 1,}\] where RR11 is the risk ratio (RR) for heavy smokers with a variant genotype, RR10 is the risk ratio for a variant genotype among nonsmokers, and RR01 is the risk ratio for heavy smokers with a nonvariant genotype. ORs were used to estimate RRs. An ICR greater than zero implies a greater than additive relationship between the genotype and smoking (interaction), whereas an ICR of zero implies an additive relationship (no interaction) and an ICR less than zero implies a less than additive relationship (negative interaction) (49). Confidence limits for ICRs that exclude zero were considered statistically significant at an alpha level of .05. For estimation of ICRs, cigarette smoking was dichotomized using the highest quartile of smoking as the exposed group. Results Characteristics of the pancreatic cancer case subjects and the control subjects are presented in Table 1. The distributions of age and sex among these two groups of subjects were approximately equal because these were variables used to frequency match the case and control subjects. Two characteristics were associated with an increased risk of pancreatic cancer: being a heavy smoker, which was defined as being in the highest quartile of smoking pack-years (≥41 pack-years) among control subjects who smoked (age- and sex-adjusted OR = 2.3, 95% CI = 1.6 to 3.3), and having 12 or fewer years of formal education (age-, sex-, and pack-years-adjusted OR = 1.5, 95% CI = 1.0 to 2.1). Age-adjusted ORs for smoking and pancreatic cancer did not differ appreciably between the women and men in our study (data not shown). The frequencies of null genotypes for GSTT1 and GSTM1, and allele and genotype frequencies for the CYP1A1 polymorphisms, by race, are presented in Table 2. The observed CYP1A1 m1 genotypes among control subjects in each racial group were in Hardy-Weinberg equilibrium (P = .12 among Caucasian participants, P = .14 among African American participants, and P = .91 among Asian participants). The observed CYP1A1 m2 genotypes among Caucasian control participants deviated from Hardy-Weinberg equilibrium (P = .003). Among Caucasians, age- and sex-adjusted ORs for risk of pancreatic cancer associated with GSTM1, GSTT1, CYP1A1 m1, and CYP1A1 m2 were near unity (Table 2). Although the CYP1A1 m4 variant (Asn461) was associated with reduced risk of pancreatic cancer among Caucasians, the risk estimate was not statistically significant. We found no statistical evidence for linkage between the CYP1A1 m2 and CYP1A1 m4 variant alleles (no concordance, data not shown). When the data were stratified by sex, ORs for GSTT1, CYP1A1 m1, and CYP1A1 m2 were higher for Caucasian women than for Caucasian men. Among the African American study participants, variant genotypes for CYP1A1 m1 were associated with a statistically nonsignificant increased risk of pancreatic cancer (Table 2). ORs for CYP1A1 m2 and CYP1A1 m4 could not be estimated among the African American subjects because of the lack of variation of these polymorphisms in this population. Among the Asian participants, increased risks of pancreatic cancer were seen for variant genotypes in CYP1A1 m1 and CYP1A1 m2; however, the numbers were small in some categories, and the 95% CIs were wide (Table 2). ORs of pancreatic cancer associated with the CYP1A1 m4 allele could not be estimated among Asians because of the lack of variation of this polymorphism in this population. We compared the combined effect of the CYP1A1 m1 and m2 polymorphisms among the Asian participants to the Asian individuals who had wild-type alleles of both genes (four case subjects, 21 control subjects). For those individuals with at least one variant allele at either locus (three case subjects, 12 control subjects), the age- and sex-adjusted OR was 1.9 (95% CI = 0.33 to 11.2); for those with at least one variant allele at both loci (10 case subjects, 19 control subjects), the age- and sex-adjusted OR was 4.1 (95% CI = 0.97 to 17.3; data not shown). GSTT1 was associated with a nearly twofold but statistically nonsignificant increased risk of pancreatic cancer among the Asian subjects (Table 2). There was no evidence among the Caucasian participants for any interactions between genotypes (data not shown). Limited sample size precluded a meaningful evaluation of interactions between GSTT1 and either CYP1A1 m2 or CYP1A1 m4 or among GSTM1, GSTT1, and CYP1A1. Stratification of genotype–genotype interaction analyses by sex did not reveal any consistent patterns. We compared the combined effect of the GSTM1 and GSTT1 genotypes among the Asian participants to the effect of the GSTM1-present/GSTT1-present genotype among Asian participants. Those participants with either the GSTM1-null/GSTT1-present or GSTM1-present/GSTT1-null genotype had an age- and sex-adjusted OR of 1.1 (95% CI = 0.24 to 5.2), whereas those with a GSTM1-null/GSTT1-null genotype had an age- and sex-adjusted OR of 2.3 (95% CI = 0.44 to 12.4; data not shown). Point estimates for main genotype effects (GSTM1, GSTT1, CYP1A1 m1, and CYP1A1 m2) among Asian participants increased when smoking duration or pack-years (but not alcohol consumption) was added to logistic models (data not shown). In general, among Asian and African American participants, estimates of ORs for combined genotypes were too imprecise for meaningful interpretation. We evaluated the data for a GSTT1-smoking interaction among Caucasian participants and found that the age- and sex-adjusted risk of pancreatic cancer was higher for those who were heavy smokers (40 years old or older or 41 pack-years or more) and had a GSTT1-null genotype than for those who were heavy smokers and had a GSTT1-present genotype (Table 3). We also calculated the ICRs to estimate the magnitude of the interaction between GSTT1-null genotype and heavy smoking. For example, for both sexes combined, the age- and sex-adjusted ICR was 2.2 (95% CI = −0.58 to 4.9) for the GSTT1-null genotype and 40 years or more of smoking and 1.9 (95% CI = −0.46 to 4.2) for the GSTT1-null genotype and 41 pack-years or more of smoking. Age-adjusted ICRs for the GSTT1-null genotype and 40 years or more of smoking by sex were 5.4 (95% CI = −2.4 to 13.2) for women and 0.64 (95% CI = −1.7 to 3.0) for men. Age-adjusted ICRs for the GSTT1-null genotype and 41 pack-years or more of smoking by sex were 2.9 (95% CI = −2.1 to 8.0) for women and 1.4 (95% CI = −1.0 to 3.9) for men. We tested for interactions between heavy smoking and GSTM1-null or GSTT1-null genotypes among Caucasian participants and found that the associations and interaction effects were weaker for the GSTM1-null genotype than for the GSTT1-null genotype. For both sexes combined, the age- and sex-adjusted OR (and ICR) for heavy smokers (≥41 pack-years) with the GSTM1-null genotype was 2.5 (95% CI = 1.4 to 4.3, ICR = 1.0; 95% CI = −0.30 to 2.3), whereas the age- and sex-adjusted OR for heavy smokers with the GSTM1-present genotype was 1.6 (95% CI = 0.93 to 2.9). The age- and sex-adjusted OR for light smokers (<41 pack-years) with the GSTM1-null genotype was 0.93 (95% CI = 0.57 to 1.5), and the age- and sex-adjusted OR for light smokers with the GSTM1-present genotype was 0.88 (95% CI = 0.53 to 1.4). The age- and sex-adjusted OR for never smokers with the GSTM1-null genotype was 0.74 (95% CI = 0.43 to 1.2). Combined ORs for heavy smoking and GSTM1 were not substantially different when the data were stratified by sex (data not shown). We then evaluated the data for a gene–gene environment interaction between GSTT1 genotype and GSTM1 genotype and smoking by estimating the OR for risk of pancreatic cancer among Caucasian subjects who were heavy smokers and had both the GSTM1-null and GSTT1-null genotypes (Table 4). These combined ORs were 4.2 (95% CI = 1.4 to 12.3) and 3.5 (95% CI = 1.4 to 8.8) for heavy smokers with 40 years or more of smoking and 41 or more pack-years of smoking, respectively, with the GSTT1-null and GSTM1-null genotypes; ORs were similar in magnitude to the combined ORs estimated for heavy smokers with the GSTT1-null genotype (both sexes combined; the OR for 40 years or more of smoking was 4.1 [95% CI = 1.9 to 8.9] and the OR for 41 or more pack-years of smoking was 3.9 [95% CI = 2.0 to 7.7]). These results suggest that the GSTM1-null genotype does not modify the observed interaction of GSTT1-null genotype with heavy smoking. However, when stratified by sex, the data on gene–gene environment interactions were too sparse for meaningful interpretation. We evaluated the data for potential interactions between CYP1A1 polymorphisms and smoking. We found no evidence for interactions between the CYP1A1 m1 variants and either smoking duration or pack-years (data not shown). Among nonsmoking women (but not men), the OR for those with CYP1A1 m1 variant genotypes was elevated almost twofold (OR = 1.8, 95% CI = 0.87 to 3.7) compared with nonsmokers with the CYP1A1 wild-type genotype. Among Caucasian participants, we found no interactions between smoking and variant genotypes for CYP1A1 m2 and CYP1A1 m4 nor among years since cessation of smoking and GSTM1, GSTT1, or CYP1A1 m1 (data not shown). Further, there was no evidence that any gene-smoking interactions were more pronounced if pack-years or duration of smoking were restricted to smoking within 15 years of the date of diagnosis for case subjects or the date of interview for control subjects (data not shown). Insufficient numbers precluded analyses of gene-smoking interactions for the GSTM1, GSTT1, and CYP1A1 polymorphisms among African American and Asian participants. We noted that, for some comparisons, the prevalence of heavy smoking (≥40 years' duration) was lower among control group subjects with a variant genotype than it was among control group subjects with a wild-type genotype (Table 3). To investigate the potential role of age in the lower prevalence of heavy smoking among subjects with variant genotypes, we analyzed the data for associations between age at diagnosis or interview and genotypes among the smoking duration categories and found none (data not shown). In addition, we found that none of the polymorphisms evaluated in this study was associated with age at diagnosis among pancreatic cancer case subjects of any racial group (data not shown) and that none of the polymorphisms was statistically significantly associated with age at interview among either the Caucasian or African American control subjects (data not shown). Among the Asian control subjects, age at interview was statistically significantly lower among the those with the GSTM1-null genotype (56.8 years) than among those with the GSTM1-present genotype (65.4 years, Wilcoxon rank sum test, P = .01). Discussion We studied the associations between genetic polymorphisms at loci that encode carcinogen-metabolizing enzymes, smoking, and pancreatic adenocarcinoma in a population-based case–control study in the San Francisco Bay area. In an analysis to detect gene-environment interactions among the Caucasian participants, we found evidence that suggests that the homozygous deletion of GSTT1 and heavy cigarette smoking has a synergistic effect on the risk of pancreatic cancer. These observations support the hypothesis that inherited deletion polymorphisms in GSTT1 increase susceptibility to smoking-related pancreatic adenocarcinoma, possibly by interfering with the detoxication of tobacco-associated carcinogens or reactive endogenous intermediates. These associations and interaction effects were stronger among women than among men, suggesting that hormones or other gender-specific factors may play a role in mediating the effects of cigarette smoking on pancreatic carcinogenesis. Consistent with our findings, phenotypic studies (36,37,50) suggest that women with GSTT1-null and GSTM1-null genotypes may be more susceptible to the effects of DNA-damaging agents than men with those genotypes. In addition, three previous case–control studies (3,4,51) and one recent prospective cohort study (52) observed higher smoking-related relative risks of pancreatic cancer among women than among men. The finding of higher relative risk among women for the combined effect of GSTT1 and smoking in our study requires replication in other population-based studies. We found limited evidence for an increased risk of pancreatic cancer among Asian participants who had the GSTT1-null genotype and variant CYP1A1 m1 and m2 alleles (alone or in combination). However, most of these associations were caused by an excess of heterozygotes among the case subjects. Although it is not clear why these associations were confined to one ethnic/racial group, there is published evidence (53) of ethnic and geographic variation in pancreatic cancer incidence, with Japanese having among the highest rates worldwide. Interestingly, eight (47%) of the 17 Asian case subjects in our study reported that both of their parents were of Japanese heritage. Thus, although we cannot rule out the possibility that the increased risk of pancreatic cancer we observed among the Asian participants was because of chance and imprecision resulting from the small number of Asian participants, we also cannot rule out the possibility that unknown disease-related alleles at linked loci may be driving the observed associations among the Asian participants. Two earlier studies (54,55) of metabolic gene polymorphisms evaluated risk of pancreatic cancer associated with main gene effects only. One study (54) investigated CYP1A1 (m1 and m2) polymorphisms and pancreatic cancer among Korean case and control subjects, whereas the other study (55) investigated GSTM1 polymorphisms in a sample of Caucasians. Neither study reported an association between these polymorphisms and pancreatic cancer. A third, more recent case–control study (56) of pancreatic adenocarcinoma evaluated risk associated with null alleles of GSTM1 and GSTT1 and with the m2 allele in CYP1A1 and found neither an overall association between these genotypes and pancreatic cancer risk nor evidence of a main effect of smoking or interactions between smoking and these genotypes. However, each of these studies had small sample sizes that precluded more detailed evaluation and interpretation of the data. The mechanisms by which carcinogens in tobacco smoke affect the pancreas are currently unknown. It is possible that these mechanisms may involve the direct actions of tobacco-associated substrates of GSTT1 on pancreatic tissues. Alternatively, carcinogens in tobacco smoke may indirectly affect GSTT1 substrates derived from smoking-associated oxidative species and/or inflammatory processes in the pancreas. The ORs and ICRs for the combined effect of heavy smoking and the GSTM1-null genotype determined in our study, plus evidence from two previous studies (55,56), suggest that detoxication of tobacco-associated PAH (e.g., BPDE) by GSTM1 is unlikely to play a major role in pancreatic carcinogenesis. Moreover, the lack of interaction between smoking and CYP1A1 polymorphisms in this study also suggests that the major mechanism by which smoking increases the risk of pancreatic cancer is not through the direct action of tobacco-associated constituents (e.g., PAH) that are modified by CYP1A1. Further, evidence for a direct effect of smoking on pancreatic cancer risk is not supported by the spectra of p53 and K-ras mutations that are commonly observed in exocrine pancreatic adenocarcinoma (57,58). The spectrum of p53 mutations in pancreatic adenocarcinoma is mixed in that the number and predominance of transversions of the p53 gene are more similar to those associated with bladder cancer and colorectal cancer than with smoking-related cancers of the lung, head and neck, and esophagus (57,58). In addition, at least 75% of pancreatic tumors show mutations in codon 12 of the K-ras oncogene. As is seen in colorectal cancer, transition mutations also predominate in pancreatic cancer, whereas in lung cancers, transversion mutations predominate (59). Our results, which showed an interaction between the GSTT1-null genotype and smoking and a lack of interaction between either the GSTM1-null genotype or variant CYP1A1 alleles and smoking, together with results published on p53 and K-ras mutations, are consistent with the hypothesis that an indirect effect via an endogenous mechanism associated with tobacco smoking is involved in pancreatic carcinogenesis. GSTT1 protects cells from the natural byproducts of lipid peroxidation and oxidative stress (e.g., hydroxyalkenals and ethylene oxide) (25), and the GSTs have been implicated in susceptibilities to other inflammatory diseases, such as hepatitis and ulcerative colitis (60–62). It is therefore possible that individuals who smoke at high intensity and for prolonged periods and who are carriers of inherited GSTT1-null alleles could have constitutively high levels of lipid peroxidation byproducts and oxidative damage in their pancreatic tissues. This damage could cause an increased risk of pancreatic cancer because of increases in mutations, chromosomal abnormalities, and cell division or tumor-clone selective pressure (63). Smoking has been associated with pancreatic tissue damage observed at autopsy (64), and DNA adducts associated with oxidative stress and phospholipid peroxidation (e.g., malondialdehyde-DNA adducts) have also been detected in human pancreata (6,7,65,66). However, because DNA adduct levels have not been consistently correlated with variables for smoking, age, sex, body mass index, or genotypes for GSTT1 and GSTM1 (65,66), it is possible that other genes involved in the oxidative stress response and DNA repair pathways may mediate the potential synergy between GSTT1 and smoking that affects pancreatic cancer risk. Furthermore, expression of GSTs in the pancreas is variable and may be influenced by other factors in addition to these genetic polymorphisms (32). We found that the potential modification of smoking-related relative risks for pancreatic cancer by GSTT1 applied to a relatively modest subset of individuals. Using our data on age- and sex-adjusted ORs for individuals with the GSTT1 genotypes and 41 pack-years or more of smoking, we estimated that 14% of pancreatic cancer cases among the Caucasian subjects with the GSTT1-present genotype and 34% of pancreatic cancer cases among the Caucasian subjects with the GSTT1-null genotype would have been prevented had those individuals decreased or discontinued their heavy smoking. However, the real importance of these findings may be to further our understanding of the etiology of this deadly cancer. A limitation of our study was the low participation rate among the pancreatic cancer case subjects who rapidly succumbed to the disease after diagnosis. Consequently, we cannot rule out the possibility that one or more of the genes examined in this study may be associated with tumor aggressiveness or disease mortality and thus may have affected case subject participation. Although all of the comparisons in our study, including stratifications by sex, were based on solid biologic rationale, our finding of increased relative risk among women, but not men, who were heavy smokers and carried a GSTT1 gene deletion, was unexpected. An alternative explanation is that the true background rates of pancreatic cancer among nonsmokers may differ by sex. Nonetheless, this finding is consistent with reports of higher smoking-related relative risks of pancreatic cancer among women than among men (3,4,51,52) and with recent reports showing that elevated relative risks for other smoking-related cancers are associated with the GSTT1 gene deletion (46,47). The strengths of our study were its population-based study design and the use of rapid case ascertainment, which allowed exposure and covariate information to be gathered from in-person interviews rather than from proxy interviews. Our data support the hypothesis that the GSTT1 enzyme protects pancreatic cells from the damaging effects of tobacco smoking and that lacking this enzyme may increase the risk of smoking-related pancreatic cancer. Furthermore, women who lack the GSTT1 enzyme may be more susceptible to these damaging effects than men who lack the enzyme. The data also suggest that the m1 and m2 alleles of CYP1A1 and the gene deletion allele for GSTT1 increase the risk of pancreatic cancer among Asian participants. The absence of consistent interaction between smoking and GSTM1-null or variant CYP1A1 genotypes does not rule out the possibility that tobacco-related carcinogens act directly in the pancreas but suggests that other environmental carcinogens are likely to play a role in the etiology of pancreatic cancer. Likewise, the possible interaction between GSTT1 gene deletion and heavy smoking in this study does not exclude the potential roles that other genes and pathways may have in the metabolism of endogenous intermediates in pancreatic carcinogenesis. Table 1. Characteristics of study participants, San Francisco Bay area, California, 1994–2001   Case subjects (n = 309)  Control subjects (n = 964)  Characteristic  No.  %  No.  %  *Unable to calculate because of missing questionnaire information.  †Not included in analyses.  Age, y      24–54  58  19  233  24      55–66  107  35  254  26      67–73  65  21  252  26      74–85  79  26  225  23  Sex      Women  141  46  433  45      Men  168  54  531  55  Ethnicity      Caucasian  261  84  860  89      African American  26  8  36  4      Asian  17  6  53  5      Other  5  2  15  2  Education      High school or less  131  42  298  31      College  119  39  423  44      Graduate school  59  19  243  25  Cigarette smoking duration, y      Never smoked  87  29  328  36      1–13  25  8  150  16      14–26  46  15  147  16      27–39  62  21  146  16      ≥40  79  26  148  16      Missing*  0    3    Cigarette smoking, pack-years      Never smoked  87  29  328  36      <6.3  27  9  143  16      6.3–20.21  43  14  146  16      20.22–40.99  55  18  150  16      ≥41  85  28  145  16      Missing*  2    10    Never smoked cigarettes, but smoked cigars and/or pipes†  10  3  42  4    Case subjects (n = 309)  Control subjects (n = 964)  Characteristic  No.  %  No.  %  *Unable to calculate because of missing questionnaire information.  †Not included in analyses.  Age, y      24–54  58  19  233  24      55–66  107  35  254  26      67–73  65  21  252  26      74–85  79  26  225  23  Sex      Women  141  46  433  45      Men  168  54  531  55  Ethnicity      Caucasian  261  84  860  89      African American  26  8  36  4      Asian  17  6  53  5      Other  5  2  15  2  Education      High school or less  131  42  298  31      College  119  39  423  44      Graduate school  59  19  243  25  Cigarette smoking duration, y      Never smoked  87  29  328  36      1–13  25  8  150  16      14–26  46  15  147  16      27–39  62  21  146  16      ≥40  79  26  148  16      Missing*  0    3    Cigarette smoking, pack-years      Never smoked  87  29  328  36      <6.3  27  9  143  16      6.3–20.21  43  14  146  16      20.22–40.99  55  18  150  16      ≥41  85  28  145  16      Missing*  2    10    Never smoked cigarettes, but smoked cigars and/or pipes†  10  3  42  4  View Large Table 2. Genotype and allele frequencies among pancreatic cancer case and control subjects, and odds ratios (ORs) and 95% confidence intervals (CIs) for main gene effects by ethnicity, San Francisco Bay area, California, 1994–2001*   Caucasians  African Americans  Asians    Case subjects (n = 261)  Control subjects (n = 860)  Case subjects (n = 26)  Control subjects (n = 36)  Case subjects (n = 17)  Control subjects (n = 53)  Genotype  No.  %  No.  %  No.  %  No.  %  No.  %  No.  %  *NE = not estimated; WT = wild type.  †Genotype data missing. All percentages, including allele frequency, were calculated by using nonmissing data.  ‡OR for null (homozygous deletion) relative to present (wild-type or heterozygous deletion); adjusted for age and sex.  §Allele frequencies were calculated by using the formula (2aa + Aa)/2N, where aa is the number of participants with a homozygous variant genotype, Aa is the number of participants with a heterozygous genotype, and 2N is the total number of chromosomes (two times the total number of participants).  ∥OR for heterozygous or homozygous variant relative to wild-type; adjusted for age and sex.  GSTM1      Present  122  47  405  47  17  65  22  61  6  35  20  38      Null  138  53  449  53  9  35  14  39  11  65  32  62      Missing†  1    6    1        OR‡ (95% CI)    1.0 (0.77 to 1.3)      0.84 (0.29 to 2.4)      1.3 (0.40 to 4.3)    GSTT1      Present  214  82  704  83  21  81  28  78  8  47  31  60      Null  46  18  149  17  5  19  8  22  9  53  21  40      Missing†  1    7    1        OR‡ (95% CI)    1.0 (0.70 to 1.4)      0.86 (0.24 to 3.1)      1.9 (0.59 to 5.9)    CYP1A1 m1      Allele frequency§          WT    88    87    83    85    56    64          M1    12    13    17    15    44    36      Genotype          WT/WT  195  76  653  77  18  69  27  75  4  24  22  42          WT/M1  62  24  180  21  7  27  7  19  11  65  24  45          M1/M1  1  <1  19  2  1  4  2  6  2  12  7  13          Missing†  3    8                OR∥ (95% CI)    1.1 (0.77 to 1.5)      1.4 (0.42 to 4.6)      3.4 (0.87 to 13.1)    CYP1A1 m2      Allele frequency§          Ile    95    94    98    100    71    80          Val    5    6    2    0    29    20      Genotype          Ile/Ile  234  90  761  89  25  96  36  100  7  41  32  62          Ile/Val  26  10  87  10  1  4  0  0  10  59  19  37          Val/Val  0  0  8  1  0  0  0  0  0  0  1  2          Missing†  1    4    1            OR∥ (95% CI)    0.90 (0.57 to 1.4)      NE      2.9 (0.86 to 9.5)    CYP1A1 m4      Allele frequency§          Thr    96    95    100    100    100    100          Asn    4    5    0    0    0    0      Genotype          Thr/Thr  240  92  775  90  26  100  36  100  17  100  53  100          Thr/Asn  19  7  81  9  0  0  0  0  0  0  0  0          Asn/Asn  1  <1  1  <1  0  0  0  0  0  0  0  0          Missing†  1    3                OR∥ (95% CI)    0.79 (0.47 to 1.3)      NE      NE      Caucasians  African Americans  Asians    Case subjects (n = 261)  Control subjects (n = 860)  Case subjects (n = 26)  Control subjects (n = 36)  Case subjects (n = 17)  Control subjects (n = 53)  Genotype  No.  %  No.  %  No.  %  No.  %  No.  %  No.  %  *NE = not estimated; WT = wild type.  †Genotype data missing. All percentages, including allele frequency, were calculated by using nonmissing data.  ‡OR for null (homozygous deletion) relative to present (wild-type or heterozygous deletion); adjusted for age and sex.  §Allele frequencies were calculated by using the formula (2aa + Aa)/2N, where aa is the number of participants with a homozygous variant genotype, Aa is the number of participants with a heterozygous genotype, and 2N is the total number of chromosomes (two times the total number of participants).  ∥OR for heterozygous or homozygous variant relative to wild-type; adjusted for age and sex.  GSTM1      Present  122  47  405  47  17  65  22  61  6  35  20  38      Null  138  53  449  53  9  35  14  39  11  65  32  62      Missing†  1    6    1        OR‡ (95% CI)    1.0 (0.77 to 1.3)      0.84 (0.29 to 2.4)      1.3 (0.40 to 4.3)    GSTT1      Present  214  82  704  83  21  81  28  78  8  47  31  60      Null  46  18  149  17  5  19  8  22  9  53  21  40      Missing†  1    7    1        OR‡ (95% CI)    1.0 (0.70 to 1.4)      0.86 (0.24 to 3.1)      1.9 (0.59 to 5.9)    CYP1A1 m1      Allele frequency§          WT    88    87    83    85    56    64          M1    12    13    17    15    44    36      Genotype          WT/WT  195  76  653  77  18  69  27  75  4  24  22  42          WT/M1  62  24  180  21  7  27  7  19  11  65  24  45          M1/M1  1  <1  19  2  1  4  2  6  2  12  7  13          Missing†  3    8                OR∥ (95% CI)    1.1 (0.77 to 1.5)      1.4 (0.42 to 4.6)      3.4 (0.87 to 13.1)    CYP1A1 m2      Allele frequency§          Ile    95    94    98    100    71    80          Val    5    6    2    0    29    20      Genotype          Ile/Ile  234  90  761  89  25  96  36  100  7  41  32  62          Ile/Val  26  10  87  10  1  4  0  0  10  59  19  37          Val/Val  0  0  8  1  0  0  0  0  0  0  1  2          Missing†  1    4    1            OR∥ (95% CI)    0.90 (0.57 to 1.4)      NE      2.9 (0.86 to 9.5)    CYP1A1 m4      Allele frequency§          Thr    96    95    100    100    100    100          Asn    4    5    0    0    0    0      Genotype          Thr/Thr  240  92  775  90  26  100  36  100  17  100  53  100          Thr/Asn  19  7  81  9  0  0  0  0  0  0  0  0          Asn/Asn  1  <1  1  <1  0  0  0  0  0  0  0  0          Missing†  1    3                OR∥ (95% CI)    0.79 (0.47 to 1.3)      NE      NE    View Large Table 3. Odds ratios (ORs) and 95% confidence intervals (CIs) for pancreatic cancer in relation to smoking and GSTT1 genotype in Caucasian participants, women and men combined and analyzed separately, San Francisco Bay area, California, 1994–2001   GSTT1-present genotype  GSTT1-null genotype  Cigarette smoking  Case subjects, No. (%)  Control subjects, No. (%)  OR (95% CI)  Case subjects, No. (%)  Control subjects, No. (%)  OR (95% CI)  *ORs adjusted for age and sex.  †Missing data on smoking. All percentages calculated by using nonmissing data.  ‡ORs adjusted for age.  Both sexes combined*  n = 207  n = 674    n = 44  n = 140        Duration, y          Never smoked  58 (28)  231 (34)  1.0 (referent)  11 (25)  46 (33)  0.95 (0.46 to 2.0)          <40  94 (45)  319 (48)  1.2 (0.81 to 1.7)  18 (41)  79 (56)  0.91 (0.51 to 1.6)          ≥40  55 (27)  121 (18)  1.8 (1.2 to 2.9)  15 (34)  15 (11)  4.1 (1.9 to 8.9)          Missing†    3              Pack-years          Never smoked  58 (28)  231 (35)  1.0 (referent)  11 (25)  46 (33)  0.95 (0.46 to 1.9)          <41  90 (44)  323 (49)  1.1 (0.78 to 1.6)  13 (30)  72 (52)  0.74 (0.38 to 1.4)          ≥41  57 (28)  111 (17)  2.1 (1.4 to 3.3)  20 (45)  21 (15)  3.9 (2.0 to 7.7)          Missing†  2  9      1    Women‡  n = 94  n = 322    n = 24  n = 63        Duration, y          Never smoked  36 (38)  141 (44)  1.0 (referent)  9 (38)  27 (43)  1.3 (0.56 to 3.0)          <40  35 (37)  124 (39)  1.1 (0.64 to 1.8)  6 (25)  31 (49)  0.75 (0.29 to 1.9)          ≥40  23 (24)  55 (17)  1.7 (0.90 to 3.1)  9 (38)  5 (8)  7.2 (2.2 to 23.1)          Missing†    2        Pack-years          Never smoked  36 (39)  141 (45)  1.0 (referent)  9 (38)  27 (44)  1.3 (0.56 to 3.0)          <41  38 (41)  138 (44)  1.1 (0.65 to 1.8)  6 (25)  28 (45)  0.84 (0.32 to 2.2)          ≥41  18 (20)  35 (11)  2.0 (1.0 to 4.0)  9 (38)  7 (11)  5.0 (1.8 to 14.5)          Missing†  2  8      1    Men‡  n = 113  n = 352    n = 20  n = 77        Duration, y          Never smoked  22 (19)  90 (26)  1.0 (referent)  2 (10)  19 (25)  0.43 (0.09 to 2.0)          <40  59 (52)  195 (56)  1.2 (0.71 to 2.1)  12 (60)  48 (62)  1.0 (0.47 to 2.3)          ≥40  32 (28)  66 (19)  2.0 (1.1 to 3.9)  6 (30)  10 (13)  2.5 (0.82 to 7.7)          Missing†    1        Pack-years          Never smoked  22 (19)  90 (26)  1.0 (referent)  2 (10)  19 (25)  0.43 (0.09 to 2.0)          <41  52 (46)  185 (53)  1.2 (0.66 to 2.0)  7 (35)  44 (57)  0.66 (0.26 to 1.6)          ≥41  39 (35)  76 (22)  2.1 (1.1 to 3.9)  11 (55)  14 (18)  3.2 (1.3 to 8.1)          Missing†    1      GSTT1-present genotype  GSTT1-null genotype  Cigarette smoking  Case subjects, No. (%)  Control subjects, No. (%)  OR (95% CI)  Case subjects, No. (%)  Control subjects, No. (%)  OR (95% CI)  *ORs adjusted for age and sex.  †Missing data on smoking. All percentages calculated by using nonmissing data.  ‡ORs adjusted for age.  Both sexes combined*  n = 207  n = 674    n = 44  n = 140        Duration, y          Never smoked  58 (28)  231 (34)  1.0 (referent)  11 (25)  46 (33)  0.95 (0.46 to 2.0)          <40  94 (45)  319 (48)  1.2 (0.81 to 1.7)  18 (41)  79 (56)  0.91 (0.51 to 1.6)          ≥40  55 (27)  121 (18)  1.8 (1.2 to 2.9)  15 (34)  15 (11)  4.1 (1.9 to 8.9)          Missing†    3              Pack-years          Never smoked  58 (28)  231 (35)  1.0 (referent)  11 (25)  46 (33)  0.95 (0.46 to 1.9)          <41  90 (44)  323 (49)  1.1 (0.78 to 1.6)  13 (30)  72 (52)  0.74 (0.38 to 1.4)          ≥41  57 (28)  111 (17)  2.1 (1.4 to 3.3)  20 (45)  21 (15)  3.9 (2.0 to 7.7)          Missing†  2  9      1    Women‡  n = 94  n = 322    n = 24  n = 63        Duration, y          Never smoked  36 (38)  141 (44)  1.0 (referent)  9 (38)  27 (43)  1.3 (0.56 to 3.0)          <40  35 (37)  124 (39)  1.1 (0.64 to 1.8)  6 (25)  31 (49)  0.75 (0.29 to 1.9)          ≥40  23 (24)  55 (17)  1.7 (0.90 to 3.1)  9 (38)  5 (8)  7.2 (2.2 to 23.1)          Missing†    2        Pack-years          Never smoked  36 (39)  141 (45)  1.0 (referent)  9 (38)  27 (44)  1.3 (0.56 to 3.0)          <41  38 (41)  138 (44)  1.1 (0.65 to 1.8)  6 (25)  28 (45)  0.84 (0.32 to 2.2)          ≥41  18 (20)  35 (11)  2.0 (1.0 to 4.0)  9 (38)  7 (11)  5.0 (1.8 to 14.5)          Missing†  2  8      1    Men‡  n = 113  n = 352    n = 20  n = 77        Duration, y          Never smoked  22 (19)  90 (26)  1.0 (referent)  2 (10)  19 (25)  0.43 (0.09 to 2.0)          <40  59 (52)  195 (56)  1.2 (0.71 to 2.1)  12 (60)  48 (62)  1.0 (0.47 to 2.3)          ≥40  32 (28)  66 (19)  2.0 (1.1 to 3.9)  6 (30)  10 (13)  2.5 (0.82 to 7.7)          Missing†    1        Pack-years          Never smoked  22 (19)  90 (26)  1.0 (referent)  2 (10)  19 (25)  0.43 (0.09 to 2.0)          <41  52 (46)  185 (53)  1.2 (0.66 to 2.0)  7 (35)  44 (57)  0.66 (0.26 to 1.6)          ≥41  39 (35)  76 (22)  2.1 (1.1 to 3.9)  11 (55)  14 (18)  3.2 (1.3 to 8.1)          Missing†    1    View Large Table 4. Odds ratios (ORs) and 95% confidence intervals (CIs) for pancreatic cancer and the combined effect of smoking and the GSTT1 and GSTM1 genotypes among Caucasian participants, San Francisco Bay area, California, 1994–2001 Cigarette smoking  GSTT1/GSTM1  No. of case subjects (n = 252)  No. of control subjects (n = 820)  OR* (95% CI)  *Adjusted for age and sex.  †Either GSTM1 or GSTT1 is null (homozygously deleted).  ‡Missing data on smoking and/or genotype.  Duration, y      Never smoked  Both present  30  96  1.0 (referent)    Either null†  32  153  0.67 (0.38 to 1.2)    Both null  7  26  0.86 (0.34 to 2.2)      <40  Both present  47  166  0.90 (0.54 to 1.5)    Either null†  53  189  0.90 (0.54 to 1.5)    Both null  12  43  0.90 (0.42 to 1.9)      ≥40  Both present  24  60  1.3 (0.69 to 2.5)    Either null†  37  69  1.8 (0.98 to 3.1)    Both null  9  7  4.2 (1.4 to 12.3)      Missing‡    1  11    Pack-years      Never smoked  Both present  30  96  1.0 (referent)    Either null†  32  153  0.66 (0.38 to 1.2)    Both null  7  26  0.85 (0.34 to 2.2)      <41  Both present  46  164  0.91 (0.54 to 1.5)    Either null†  48  192  0.81 (0.48 to 1.4)    Both null  9  39  0.75 (0.33 to 1.7)      ≥41  Both present  24  58  1.4 (0.72 to 2.6)    Either null†  41  63  2.2 (1.2 to 3.9)    Both null  12  11  3.5 (1.4 to 8.8)      Missing‡    3  18    Cigarette smoking  GSTT1/GSTM1  No. of case subjects (n = 252)  No. of control subjects (n = 820)  OR* (95% CI)  *Adjusted for age and sex.  †Either GSTM1 or GSTT1 is null (homozygously deleted).  ‡Missing data on smoking and/or genotype.  Duration, y      Never smoked  Both present  30  96  1.0 (referent)    Either null†  32  153  0.67 (0.38 to 1.2)    Both null  7  26  0.86 (0.34 to 2.2)      <40  Both present  47  166  0.90 (0.54 to 1.5)    Either null†  53  189  0.90 (0.54 to 1.5)    Both null  12  43  0.90 (0.42 to 1.9)      ≥40  Both present  24  60  1.3 (0.69 to 2.5)    Either null†  37  69  1.8 (0.98 to 3.1)    Both null  9  7  4.2 (1.4 to 12.3)      Missing‡    1  11    Pack-years      Never smoked  Both present  30  96  1.0 (referent)    Either null†  32  153  0.66 (0.38 to 1.2)    Both null  7  26  0.85 (0.34 to 2.2)      <41  Both present  46  164  0.91 (0.54 to 1.5)    Either null†  48  192  0.81 (0.48 to 1.4)    Both null  9  39  0.75 (0.33 to 1.7)      ≥41  Both present  24  58  1.4 (0.72 to 2.6)    Either null†  41  63  2.2 (1.2 to 3.9)    Both null  12  11  3.5 (1.4 to 8.8)      Missing‡    3  18    View Large Present address: E. J. Duell, Department of Epidemiology and Social Medicine, Albert Einstein College of Medicine, Bronx, NY 10461. Supported by Public Health Service (PHS) grants CA59706 and CA89726 to E. A. Holly from the National Cancer Institute (NCI), National Institutes of Health (NIH), Department of Health and Human Services (DHHS); PHS grant ES00002 from the National Institute of Environmental Health Sciences (NIEHS), NIH, DHHS; and PHS grant CA78609 from the NCI, NIH, DHHS to K. T. Kelsey; by ES06717 from NIEHS to J. K. Wiencke; by PHS grant CA09078 from the NCI to E. J. Duell; and by the Lustgarten Foundation for Pancreatic Cancer Research to E. J. Duell. We thank Drs. Tomoko Hirao and Duk-Hwan Kim for assistance with genotyping. References 1 Greenlee RT, Hill-Harmon MB, Murray T, Thun M. Cancer statistics, 2001. A Cancer J Clin  2001; 51: 15–36. Google Scholar 2 Weiderpass E, Partanen T, Kaaks R, Vainio H, Porta M, Kauppinen T, et al. Occurrence, trends and environmental etiology of pancreatic cancer. Scand J Work Environ Health  1998; 24: 165–74. Google Scholar 3 Howe GR, Jain M, Burch JD, Miller AB. Cigarette smoking and cancer of the pancreas: evidence from a population-based case–control study in Toronto, Canada. Int J Cancer  1991; 47: 323–8. Google Scholar 4 Silverman DT, Dunn JA, Hoover RN, Schiffman M, Lillemoe KD, Schoenberg JB, et al. Cigarette smoking and pancreas cancer: a case–control study based on direct interviews. J Natl Cancer Inst  1994; 86: 1510–16. Google Scholar 5 Fuchs CS, Colditz GA, Stampfer MJ, Giovannucci EL, Hunter DJ, Rimm EB, et al. A prospective study of cigarette smoking and the risk of pancreatic cancer. Arch Intern Med  1996; 156: 2255–60. Google Scholar 6 Anderson KE, Hammons GJ, Kadlubar FF, Potter JD, Kaderlik KR, Ilett KF, et al. Metabolic activation of aromatic amines by human pancreas. Carcinogenesis  1997; 18: 1085–92. Google Scholar 7 Wang M, Abbruzzese JL, Friess H, Hittelman WN, Evans DB, Abbruzzese MC, et al. DNA adducts in human pancreatic tissues and their potential role in carcinogenesis. Cancer Res  1998; 58: 38–41. Google Scholar 8 Talamini G, Bassi C, Falconi M, Sartori N, Salvia R, Rigo L, et al. Alcohol and smoking as risk factors in chronic pancreatitis and pancreatic cancer. Dig Dis Sci  1999; 44: 1303–11. Google Scholar 9 Kleeff J, Kusama T, Rossi DL, Ishiwata T, Maruyama H, Friess H, et al. Detection and localization of Mip-3alpha/LARC/Exodus, a macrophage proinflammatory chemokine, and its CCR6 receptor in human pancreatic cancer. Int J Cancer  1999; 81: 650–7. Google Scholar 10 Chow CK. Cigarette smoking and oxidative damage in the lung. Ann NY Acad Sci  1993; 686: 289–98. Google Scholar 11 Morrow JD, Frei B, Longmire AW, Gaziano JM, Lynch SM, Shyr Y, et al. Increase in circulating products of lipid peroxidation (F2-isoprostanes) in smokers. N Engl J Med  1995; 332: 1198–203. Google Scholar 12 Garner RC, Cuzick J, Jenkins D, Phillips DH, Hewer A, King MM, et al. Linear relationship between DNA adducts in human lung and cigarette smoking. IARC Sci Publ  1990; 104: 421–6. Google Scholar 13 Banerjee KK, Marimuthu P, Sarkar A, Chaudhuri RN. Influence of cigarette smoking on Vitamin C, glutathione and lipid peroxidation status. Indian J Public Health  1998; 42: 20–3. Google Scholar 14 Berger DH, Chang H, Wood M, Huang L, Heath CW, Lehman T, et al. Mutational activation of K-ras in nonneoplastic exocrine pancreatic lesions in relation to cigarette smoking status. Cancer  1999; 85: 326–32. Google Scholar 15 Schoenberg MH, Birk D, Beger HG. Oxidative stress in acute and chronic pancreatitis. Am J Clin Nutr  1995; 62(6 Suppl): 1306S–14S. Google Scholar 16 Talamini G, Bassi C, Falconi M, Frulloni L, Di Francesco V, Vaona B, et al. Cigarette smoking: an independent risk factor in alcoholic pancreatitis. Pancreas  1996; 12: 131–7. Google Scholar 17 Lowenfels AB, Maisonneuve P, Cavallini G, Ammann RW, Lankisch PG, Andersen JR, et al. International Pancreatitis Study Group. Pancreatitis and the risk of pancreatic cancer. N Eng J Med  1993; 328: 1433–7. Google Scholar 18 Hayashi S, Watanabe J, Nakachi K, Kawajiri K. Genetic linkage of lung cancer-associated MspI polymorphisms with amino acid replacement in the heme binding region of the human cytochrome P450IA1 gene. J Biochem  1991; 110: 407–11. Google Scholar 19 Crofts F, Cosma GN, Currie D, Taioli E, Toniolo P, Garte SJ. A novel CYP1A1 gene polymorphism in African-Americans. Carcinogenesis  1993; 14: 1729–31. Google Scholar 20 Cascorbi I, Brockmoller J, Roots I. A C4887A polymorphism in exon 7 of human CYP1A1: population frequency, mutation linkages, and impact on lung cancer susceptibility. Cancer Res  1996; 56: 4965–9. Google Scholar 21 Crofts F, Taioli E, Trachman J, Cosma GN, Currie D, Toniolo P, et al. Functional significance of different human CYP1A1 genotypes. Carcinogenesis  1994; 15: 2961–3. Google Scholar 22 Cosma G, Crofts F, Taioli E, Toniolo P, Garte S. Relationship between genotype and function of the human CYP1A1 gene. J Toxicol Environ Health  1993; 40: 309–16. Google Scholar 23 Bartsch H, Nair U, Risch A, Rojas M, Wikman H, Alexandrov K. Genetic polymorphism of CYP genes, alone or in combination, as a risk modifier of tobacco-related cancers. Cancer Epidemiol Biomarkers Prev  2000; 9: 3–28. Google Scholar 24 Foster JR, Idle JR, Hardwick JP, Bars R, Scott P, Braganza JM. Induction of drug-metabolizing enzymes in human pancreatic cancer and chronic pancreatitis. J Pathol  1993; 169: 457–63. Google Scholar 25 Ketterer B. Glutathione S-transferases and prevention of cellular free radical damage. Free Radic Res  1998; 28: 647–58. Google Scholar 26 Campbell JA, Corrigall AV, Guy A, Kirsch RE. Immunohistologic localization of alpha, mu, and pi class glutathione S-transferases in human tissues. Cancer  1991; 67: 1608–13. Google Scholar 27 Awasthi YC, Sharma R, Singhal SS. Human glutathione S-transferases. Int J Biochem  1994; 26: 295–308. Google Scholar 28 Juronen E, Tasa G, Uuskula M, Pooga M, Mikelsaar AV. Purification, characterization and tissue distribution of human class theta glutathione S-transferase T1–1. Biochem Mol Biol Int  1996; 39: 21–9. Google Scholar 29 Sherratt PJ, Pulford DJ, Harrison DJ, Green T, Hayes JD. Evidence that human class Theta glutathione S-transferase T1–1 can catalyse the activation of dichloromethane, a liver and lung carcinogen in the mouse. Comparison of the tissue distribution of GST T1–1 with that of classes Alpha, Mu and Pi GST in human. Biochem J  1997; 326: 837–46. Google Scholar 30 Collier JD, Bennett MK, Hall A, Cattan AR, Lendrum R, Bassendine MF. Expression of glutathione S-transferases in normal and malignant pancreas: an immunohistochemical study. Gut  1994; 35: 266–9. Google Scholar 31 Wallig MA, Kingston S, Staack R, Jefferey EH. Induction of rat pancreatic glutathione S-transferase and quinone reductase activities by a mixture of glucosinolate breakdown derivatives found in Brussels sprouts. Food Chem Toxicol  1998; 36: 365–73. Google Scholar 32 Coles BF, Anderson KE, Doerge DR, Churchwell MI, Lang NP, Kadlubar FF. Quantitative analysis of interindividual variation of glutathione S-transferase expression in human pancreas and the ambiguity of correlating genotype with phenotype. Cancer Res  2000; 60: 573–9. Google Scholar 33 Kelsey KT, Spitz MR, Zuo ZF, Wiencke JK. Polymorphisms in the glutathione S-transferase class mu and theta genes interact and increase susceptibility to lung cancer in minority populations (Texas, United States). Cancer Causes Control  1997; 8: 554–9. Google Scholar 34 Nelson HH, Wiencke JK, Christiani DC, Cheng TJ, Zuo ZF, Schwartz BS, et al. Ethnic differences in the prevalence of the homozygous deleted genotype of glutathione S-transferase theta. Carcinogenesis  1995; 16: 1243–5. Google Scholar 35 Pemble S, Schroeder KR, Spencer SR, Meyer DJ, Hallier E, Bolt HM, et al. Human glutathione S-transferase theta (GSTT1): cDNA cloning and the characterization of a genetic polymorphism. Biochem J  1994; 300: 271–6. Google Scholar 36 Wiencke JK, Kelsey KT, Lamela RA, Toscano WA Jr. Human glutathione S-transferase deficiency as a marker of susceptibility to epoxide-induced cytogenetic damage. Cancer Res  1990; 50: 1585–90. Google Scholar 37 Wiencke JK, Pemble S, Ketterer B, Kelsey KT. Gene deletion of glutathione S-transferase theta: correlation with induced genetic damage and potential role in endogenous mutagenesis. Cancer Epidemiol Biomarkers Prev  1995; 4: 253–9. Google Scholar 38 Rojas M, Cascorbi I, Alexandrov K, Kriek E, Auburtin G, Mayer L, et al. Modulation of benzo[a]pyrene diolepoxide-DNA adduct levels in human white blood cells by CYP1A1, GSTM1 and GSTT1 polymorphism. Carcinogenesis  2000; 21: 35–41. Google Scholar 39 Rojas M, Alexandrov K, Cascorbi I, Brockmoller J, Likhachev A, Pozharisski K, et al. High benzo[a]pyrene diol-epoxide DNA adduct levels in lung and blood cells from individuals with combined CYP1A1 MspI/Msp-GSTM1*0/*0 genotypes. Pharmacogenetics  1998; 8: 109–18. Google Scholar 40 Shields PG, Bowman ED, Harrington AM, Doan VT, Weston A. Polycyclic aromatic hydrocarbon-DNA adducts in human lung and cancer susceptibility genes. Cancer Res  1993; 53: 3486–92. Google Scholar 41 Seidegard J, Pero RW, Markowitz MM, Roush G, Miller DG, Beattie EJ. Isoenzyme(s) of glutathione transferase (class Mu) as a marker for the susceptibility to lung cancer: a follow up study. Carcinogenesis  1990; 11: 33–6. Google Scholar 42 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. Google Scholar 43 Deakin M, Elder J, Hendrickse C, Peckham D, Baldwin D, Pantin C, et al. Glutathione S-transferase GSTT1 genotypes and susceptibility to cancer: studies of interactions with GSTM1 in lung, oral, gastric and colorectal cancers. Carcinogenesis  1996; 17: 881–4. Google Scholar 44 Chenevix-Trench G, Young J, Coggan M, Board P. Glutathione S-transferase M1 and T1 polymorphisms: susceptibility to colon cancer and age of onset. Carcinogenesis  1995; 16: 1655–7. Google Scholar 45 Strange RC, Lear JT, Fryer AA. Glutathione S-transferase polymorphisms: influence on susceptibility to cancer. Chem Biol Interact  1998; 111–112: 351–64. Google Scholar 46 Olshan AF, Weissler MC, Watson MA, Bell DA. GSTM1, GSTT1, GSTP1, CYP1A1, and NAT1 polymorphisms, tobacco use, and the risk of head and neck cancer. Cancer Epidemiol Biomarkers Prev  2000; 9: 185–91. Google Scholar 47 Hou SM, Falt S, Nyberg F. Glutathione S-transferase T1-null genotype interacts synergistically with heavy smoking on lung cancer risk. Environ Mol Mutagen  2001; 38: 83–6. Google Scholar 48 Lundberg M, Fredlund P, Hallqvist J, Diderichsen F. A SAS program calculating three measures of interaction with confidence intervals. Epidemiology  1996; 7: 655–6. Google Scholar 49 Rothman KJ, Greenland S, editors. Modern epidemiology. 2nd ed. Philadelphia (PA): Lippincott; 1998. Google Scholar 50 Wiencke JK, Wrensch MR, Miike R, Petrakis NL. Individual susceptibility to induced chromosome damage and its implications for detecting genotoxic exposures in human populations. Cancer Res  1991; 51: 5266–9. Google Scholar 51 Muscat JE, Stellman SD, Hoffmann D, Wynder EL. Smoking and pancreatic cancer in men and women. Cancer Epidemiol Biomarkers Prev  1997; 6: 15–9. Google Scholar 52 Nilsen TI, Vatten LJ. A prospective study of lifestyle factors and the risk of pancreatic cancer in Nord-Trondelag, Norway. Cancer Causes Control  2000; 11: 645–52. Google Scholar 53 Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J, editors. Cancer incidence in five continents. Vol. 7. Lyon (France): International Agency for Research on Cancer; 1997 IARC Sci Publ No. 143. Google Scholar 54 Lee HC, Yoon YB, Kim CY. Association between genetic polymorphisms of the cytochromes p-450 (1A1, 2D6, and 2E1) and the susceptibility to pancreatic cancer. Korean J Intern Med  1997; 12: 128–36. Google Scholar 55 Bartsch H, Malaveille C, Lowenfels AB, Maisonneuve P, Hautefeuille A, Boyle P. Genetic polymorphism of N-acetyltransferases, glutathione S-transferase M1 and NAD(P)H:quinone oxidoreductase in relation to malignant and benign pancreatic disease risk. The International Pancreatic Disease Study Group. Eur J Cancer Prev  1998; 7: 215–23. Google Scholar 56 Liu G, Ghadirian P, Vesprini D, Hamel N, Paradis AJ, Lal G, et al. Polymorphisms in GSTM1, GSTT1 and CYP1A1 and risk of pancreatic adenocarcinoma. Br J Cancer  2000; 82: 1646–9. Google Scholar 57 Blanck HM, Tolbert PE, Hoppin JA. Patterns of genetic alterations in pancreatic cancer: a pooled analysis. Environ Mol Mutagen  1999; 33: 111–22. Google Scholar 58 Sirivatanauksorn V, Sirivatanauksorn Y, Lemoine NR. Molecular pattern of ductal pancreatic cancer. Langenbecks Arch Surg  1998; 383: 105–15. Google Scholar 59 McCormick CS, Lemoine NR. Molecular biological events in the development of pancreatic cancer. In: Beger HG, Warshaw AL, Buchler MW, Carr-Locke DL, Neoptolemos JP, Russell C, and Sarr MG, editors. The pancreas. Vol 2. Oxford (UK): Blackwell Science; 1998. p. 907–21. Google Scholar 60 Harada S, Abei M, Tanaka N, Agarwal DP, Goedde HW. Liver glutathione S-transferase polymorphism in Japanese and its pharmacogenetic importance. Hum Genet  1987; 75: 322–5. Google Scholar 61 Hertervig E, Nilsson A, Seidegard J. The expression of glutathione transferase mu in patients with inflammatory bowel disease. Scand J Gastroenterol  1994; 29: 729–35. Google Scholar 62 Duncan H, Swan C, Green J, Jones P, Brannigan K, Alldersea J, et al. Susceptibility to ulcerative colitis and Crohn's disease: interactions between glutathione S-transferase GSTM1 and GSTT1 genotypes. Clin Chim Acta  1995; 240: 53–61. Google Scholar 63 Marnett LJ, Plastaras JP. Endogenous DNA damage and mutation. Trends Genet  2001; 17: 214–21. Google Scholar 64 Auerbach O, Garfinkel L. Histological changes in pancreas in relation to smoking and coffee-drinking habits. Dig Dis Sci  1986; 31: 1014–20. Google Scholar 65 Kadlubar FF, Anderson KE, Haussermann S, Lang NP, Barone GW, Thompson PA, et al. Comparison of DNA adduct levels associated with oxidative stress in human pancreas. Mutat Res  1998; 405: 125–33. Google Scholar 66 Thompson PA, Seyedi F, Lang NP, MacLeod SL, Wogan GN, Anderson KE, et al. Comparison of DNA adduct levels associated with exogenous and endogenous exposures in human pancreas in relation to metabolic genotype. Mutat Res  1999; 424: 263–74. Google Scholar © Oxford University Press
TI - A Population-Based, Case–Control Study of Polymorphisms in Carcinogen-Metabolizing Genes, Smoking, and Pancreatic Adenocarcinoma Risk
JF - JNCI: Journal of the National Cancer Institute
DO - 10.1093/jnci/94.4.297
DA - 2002-02-20
UR - https://www.deepdyve.com/lp/oxford-university-press/a-population-based-case-control-study-of-polymorphisms-in-carcinogen-E5fsh0y472
SP - 297
EP - 306
VL - 94
IS - 4
DP - DeepDyve
ER -