Abstract
Heavy alcohol consumption, particularly in combination with cigarette smoking, increases the risk of oral squamous cell carcinoma (OSCC). Alcohol dehydrogenase 3 (ADH3) converts ethanol to acetaldehyde, which is a suspected oral carcinogen. The ADH3*1 allele is associated with increased conversion of ethanol to acetaldehyde, but whether the risk of OSCC is increased among ADH3*1 carriers, or whether the risk of OSCC attributable to alcohol consumption is modified by ADH3 genotype is unclear from previous studies. We examined the association between ADH3 genotypes, alcohol consumption, and OSCC risk in a population-based study of 333 cases and 541 controls from the state of Washington. The distribution of ADH3 genotypes was similar among cases and controls: ADH3*1/*1: 32.7% cases, 36.5% controls; ADH3*1/*2: 49.0% cases, 43.1% controls: ADH3*2/*2: 18.3% cases, 20.3% controls. The age-, sex-, and race-adjusted odds ratios (OR), relative to ADH3*2/*2 carriers, were as follows: ADH*1/*1: OR, 1.0 [95% confidence interval (CI) = 0.7, 1.5]; and ADH3*1/*2: OR, 1.3 (95% CI = 1.0, 1.8). We modeled the risk of OSCC associated with alcohol consumption as modified by ADH3 genotype adjusting for age, sex, race, and cigarette smoking. Among ADH3*2 homozygotes, the risk of OSCC increased 5.3% (2.1–8.5%) with each additional alcoholic drink/week, compared with 2.5% (1.5–2.6%) and 1.2% (0.0–2.4%) among persons carrying the ADH3*1/*2 and ADH3*1/*1 genotypes, respectively. These data suggest that the ADH3*2 allele confers increased susceptibility to the effect of alcohol on OSCC risk in our population.
Introduction
OSCC3 will be diagnosed in more than 30,000 people in the United States in the year 2001 and will result in nearly 8,000 deaths (1). At least 75% of OSCC in the United States can be attributed to cigarette smoking, heavy alcohol consumption, or the combination of these habits (2). Nonetheless, the vast majority of persons with these habits do not develop OSCC. Studies attempting to identify specific genetic factors that increase the risk of OSCC among exposed populations to date have focused primarily on cigarette smoking and the variation in genes involved in tobacco carcinogen metabolism (3).
One possible mechanism through which alcohol consumption may influence the risk of OSCC is via conversion of ethanol to acetaldehyde, an animal upper respiratory carcinogen (4). The major enzymes involved in first-pass ethanol metabolism are the ADHs, which convert ethanol to acetaldehyde, and the ALDHs, which convert acetaldehyde to acetic acid. Polymorphisms in genes for ADH2,4 ADH3, and ALDH2 have been extensively characterized (7, 8), but only the ADH3 polymorphism (Arg271→Gln and Ile349→Val) is sufficiently common to be amenable to study in Caucasian populations. The ADH3*1 allele (Arg271 and Ile349) is associated with a 2–3-fold increased rate of conversion of ethanol to acetaldehyde compared with the ADH3*2 allele (7, 8). A previous study in Puerto Rico (9) found no overall association of the ADH3*1 allele with OSCC risk, but the effect of alcohol consumption on OSCC was stronger among ADH3*1 homozygotes than among persons with other ADH3 genotypes. A second study (10) suggested that, among French alcoholics, homozygosity for ADH3*1 was associated with a >2-fold increased risk of OSCC, and that the combination of homozygosity for ADH3*1 and the GSTM1 null genotype was associated with a 4-fold OSCC risk compared with GSTM1 carriers who possessed at least one ADH3*2 allele. In contrast, two studies in the United States (11, 12) and one in France (13) found little or no increased risk of OSCC associated with homozygosity for ADH3*1 and no evidence of a modifying effect on the association with alcohol consumption.
The previous studies were either relatively small (fewer than 150 cases and 200 controls; Refs. 9, 10, 11 and 13), were conducted in populations with large proportions of heavy drinkers of alcoholic beverages (10, 13), or lacked quantitative information on alcohol consumption (12). We conducted a large population-based case-control study to examine further the relationship between alcohol use, ADH3 genotypes, and OSCC risk in a setting where heavy alcohol consumption is relatively uncommon.
Materials and Methods
Study Population
This investigation is based on participants, data, and biological specimens assembled during two previously conducted population-based case-control studies of OSCC (14, 15). These studies were originally designed to examine the association between human papillomavirus infection and OSCC risk. All participants were residents of King, Pierce, or Snohomish counties in Washington state. Eligible cases in the first study were all men 18–65 years of age diagnosed with new, incident OSCC between January 1985 and December 1989. Eligible cases in the second study included all men and women 18–65 years of age diagnosed with new, incident OSCC between January 1990 and June 1995. OSCC patients were ascertained through the population-based Cancer Surveillance System, a participant in the Surveillance, Epidemiology, and End Results program (16). In both studies OSCC was defined as in situ and invasive tumors of the tongue, gums, floor of mouth, tonsils, oropharynx, and other intraoral sites. In the earliest of these studies (14), the definition of OSCC also included cancers of the lip (exclusive of the vermilion border); these cases of lip cancer were excluded from the present investigation. Controls were identified for both previous studies through random-digit telephone dialing (17) and then frequency matched to the cases on age and sex. The protocols for recruitment of cases and controls in both studies were approved by the institutional review board of the Fred Hutchinson Cancer Research Center, Seattle, Washington.
Combining across the two studies, 407 eligible cases and 615 eligible controls had been recruited to participate in an in-person interview (see below). The participation rates for cases in the first and second study were 54.4% and 63.3%, respectively. Among the 275 nonparticipating cases in both studies, 125 had died before recruitment. The control response rates for the two studies were 63% and 61%, respectively; these rates incorporate both the household screening phase of random-digit dialing as well as the success at interviewing eligible controls among screened households.
Data and Biological Specimen Collection
Participating cases and controls were interviewed in-person by trained personnel using a structured instrument. The same instrument was used in both studies. The interview elicited demographic characteristics and extensive histories of tobacco use (cigarette smoking and smokeless tobacco chewing) and alcohol consumption as previously described (14, 15). Dietary intake data were collected from cases (n = 195) and controls (n = 408) recruited in the latter two-thirds of the second previous study (15) using a self-administered FFQ that was modified from the Block instrument (18). These figures represent 93.3% of the cases and 94.2% of the controls who were asked to complete the FFQ, but only 58.6% and 75.4% of the total number of cases and controls, respectively, included in this report. We excluded FFQ data from 16 cases and 14 controls for whom the reported calculated daily dietary intake was either <500 calories or >4500 calories.
Biological specimens from which genomic DNA could be extracted had been sought from each participating case and control during the original studies. All participating cases and controls were asked to provide a sample of exfoliated oral tissue as described previously (14, 15). In addition, from a sample of the participants in the second study, we had sought venous blood specimens that were processed into buffy coat aliquots. Finally, for OSCC cases who had been interviewed in the second study, we had available DNA that had been extracted from archival tumor tissue that had been processed into 10-μm thick sections without microdissection. Across the two studies, one or more of these specimens were available for 365 of 407 interviewed cases (89.7%) and 576 of 615 interviewed controls (93.7%). To assess the possibility of bias caused by associations between genotypes and prognosis from OSCC, as part of the present study, we also obtained archival specimens for 35 OSCC cases (29 of whom were Caucasian) who were eligible for the second of the two original studies (15) but had not participated because of they died before being recruited.
Laboratory Procedures
DNA Extraction.
Buffy coat specimens were thawed at 4°C and rinsed twice with sterile 1× PBS. RBCs were lysed in a solution of 0.1 m Tris, 0.05 m MgCl2, and 0.1 m NaCl (pH 7.6). The remaining lymphocytes were pelleted by centrifugation and resuspended in 950 μl of digestion buffer [10 mm Tris (pH 7.5), 0.3 m NaCl, 5 mm EDTA, 0.5% SDS, and 1 mg/ml proteinase K) and incubated at 65°C for 16 h. The protein was precipitated with 0.5 ml of saturated NaCl solution and pelleted by centrifugation. The supernatant was transferred to a new tube and the DNA precipitated with 2.5 volumes of 100% ethanol. Exfoliated oral tissue specimens were thawed at 4°C and digested in 1.6 mg/ml of proteinase K at 65°C for 16 h. The protein was precipitated with 0.25 ml of saturated NaCl solution and pelleted by centrifugation. The samples were purified further by phenol:chloroform (1:1) extraction, and the DNA was precipitated with 2.5 volumes of 100% ethanol. For the 35 OSCC cases who had died (see above), 10-μm thick sections from paraffin-embedded specimens were prepared from microdissected, histologically normal tissue that was obtained specifically for this genotyping study. DNA was extracted from all paraffin sections using a phenol:chloroform method (19). For all tissue specimens, the precipitated DNA was collected by centrifugation, washed twice with 70% ethanol and resuspended in TE buffer [10 mm Tris-HCl and 1 mm EDTA (pH 8.0)] and stored at −20°C until analysis.
Genotyping of ADH3 and GSTM1 Genes.
A 145-bp fragment of ADH3 was amplified using primers 5′ GCT TTA AGA GTA AAT ATT CTG TCC CC 3′ and 5′ AAT CTA CCT CTT TCC GAA GC 3′ (20). Each 30-μl reaction contained 15 μl of Qiagen Taq PCR Master Mix (1.5 mm MgCl2, 200 μm each dNTP, and 1.5 units Taq DNA polymerase), 0.2 μm each primer, and 100 ng of genomic DNA. All reactions were performed in a PTC-100 Thermocycler (MJ Research). Cycles were as follows: 1 cycle at 94°C for 5 min; 35 cycles at 92°C for 2 min; 53°C for 1 min; 72°C for 1 min; and then 1 cycle at 72°C for 2 min. Ten μL of the resulting PCR fragment was digested with 5 units of SspI (New England Biolabs). The fragments were separated on a 4% Nusieve gel, stained with ethidium bromide, illuminated with UV light, and photographed with Polaroid type 667 film.
We assayed for the GSTM1 null polymorphism, including coamplification of a β-globin sequence as an internal standard, using previously described methods (21). Specimens that lacked both the expected 215-bp band representing GSTM1 and a band for the β-globin fragment were considered “unscorable.”
Quality control samples included those with known genotype (positive controls) and those with PCR reagent only (negative controls). Two reviewers independently read the gels. Genotyping and genotype assignments were conducted without knowledge of the case-control status or other characteristics of each participant. ADH3 results were available for 333 cases and 541 controls, and GSTM1 results were available for 337 cases and 541 controls. Results for both genotypes were available for 328 cases and 535 controls. Among interviewed OSCC cases for whom ADH3 and GSTM1 genotyping was successful, the results were based on DNA extracted from peripheral leukocytes or exfoliated oral tissue (as opposed to archival tumor specimens) for 96.7% (ADH3) and 96.2% (GSTM1) of patients.
Statistical Analysis
We created analytic variables to describe each case and control subject’s cigarette smoking status as follows: status as of reference date (current smoker, former smoker, or never smoker), recency of cigarette smoking (among former cigarette smokers), total years smoked cigarettes, and total pack-years of cigarette smoking. Similarly, we classified case and control subjects according to their consumption of alcoholic beverages: status as of reference date (current, former, or never), and lifetime average number of alcoholic beverages consumed per week.
We computed associations between OSCC and risk factors, including ADH3 genotype, by calculating OR estimates using unconditional logistic regression models (22). ORs for ADH3 genotypes were computed with the ADH3*2/*2 carriers as the reference category. To estimate the extent to which ADH3 genotypes modify the association between alcohol consumption and oral cancer, we fit unconditional logistic regression models containing terms for ADH3 genotype, average weekly alcohol consumption, and the product of ADH3 genotype and average weekly alcohol consumption, as well as terms for age (continuous), sex, race (white or nonwhite), and cigarette smoking (continuous pack-years). In initial models to explore patterns of effect modification, we used indicator terms for both ADH3 genotype and alcohol consumption (<1 drink/week, 1–14 drinks/week, 15–28 drinks/week, ≥29 drinks/week). As these models were consistent with a codominant modifying effect of ADH3 alleles on the monotonic association between alcohol use and OSCC, subsequent models included ADH3 genotype and alcohol use as continuous terms. In analyses of continuous alcohol use, we truncated consumption at 200 drinks/week to reduce the influence of potentially erroneous reported or calculated intake (9). There were 3 cases and no controls for whom alcohol consumption data were truncated. We used hierarchical likelihood ratio tests to assess the extent to which model fits improved with different codings of ADH3 genotype and with the addition of product terms. Linear combinations of the estimated coefficients yielded the average percentage of increase in OSCC risk, and 95% CIs, associated with each additional alcoholic drink/week among persons with each ADH3 genotype. We plotted the fitted ORs for each case and control as a function of his or her weekly alcohol consumption and ADH3 genotype.
We conducted subanalyses in which we excluded nonwhite cases and controls, as well as estimated associations separately for intraoral and for pharyngeal cancers (tonsillar and oropharyngeal). We also conducted subanalyses stratified by cigarette smoking (<20, ≥20 pack-years) to examine joint associations involving ADH3 genotypes, alcohol use, and tobacco consumption. To determine whether persons carrying both the GSTM1 null genotype and the ADH3*1/*1 genotype are at increased risk of OSCC, as suggested by Coutelle et al. (10), we classified cases and controls into three groups: those carrying both the GSTM1 null and ADH3*1/*1 genotypes; those not carrying either GSTM1 null or ADH*1/*1 genotypes (the reference group for comparison); and all other combinations of the genotypes at these two loci. These analyses were stratified by alcohol consumption, because the original report by Coutelle et al. (10) was conducted among alcoholics.
Separately for cases and controls, we also compared the distribution of risk factors between participants for whom we did and did not have ADH3 genotyping results to identify any differences that may suggest the presence of bias. We also compared the distribution of ADH3 genotypes among recruited cases to the genotype distribution of cases who had died before recruitment and for whom we had genotyping results from DNA extracted from archival tumor tissue specimens. Finally, because it is possible that consumption of dietary fruits and vegetables could be an important confounder in analyses of the association between alcohol consumption and OSCC, and because we lacked dietary data on a large proportion of cases and controls in this study, among controls for whom dietary data were available we calculated partial correlations and used ordinary least squares regression to estimate associations between dietary intake of fruits and vegetables and alcohol consumption. These analyses permitted us to assess qualitatively the extent to which lack of control of confounding by fruit and vegetable consumption could have affected our results.
Results
Of the 333 cases with genotyping results for ADH3, 42.3% (n = 141) had cancers of the tongue, 22.8% (n = 76) had cancers of the tonsils or oropharynx, 15.0% (n = 50) had cancers of the floor of the mouth, 4.8% (n = 16) had cancers of the gum, 3.9% (n = 13) had cancers of the soft palate, and the remainder (n = 37) had cancers of miscellaneous intraoral sites. Characteristics of cases and controls are shown in Table 1. Cases were less likely than controls to have completed college. Heavy cigarette smoking and alcohol consumption both were strongly associated with OSCC risk, and the joint relationship exceeded that expected based on the multiplicative model. The vast majority of alcohol consumption was in the form of beer or hard liquor; whereas 35%–37% of cases and 10%–15% of controls reported an average of at least daily consumption of beer or hard liquor, respectively, the corresponding proportions for wine were <3% for cases and <2% for controls. Smokeless tobacco use was only weakly, and nonsignificantly, associated with OSCC.
Genotyped cases and controls did not differ materially from the respective nongenotyped cases and controls with respect to the characteristics shown in Table 1 (data not shown). However, genotyped cases were more likely to have tongue cancers (47.3% versus 35.1%) and less likely to have tonsillar or oropharyngeal cancers (22.8% versus 31.1%).
Among white controls, there was a slight deficit of ADH3 heterozygotes compared with that expected under the Hardy-Weinberg equilibrium; the observed proportions with ADH3*1/*1, ADH3*1/*2, and ADH3*2/*2 genotypes were 34.6%, 44.6%, and 20.6%, respectively, compared with 32.4%, 49.0%, and 18.5% expected (P = 0.042). ADH3 allele and genotype frequencies were similar between cases and controls, and there was no age-, race-, and sex-adjusted association between carriers of ADH3*1 and OSCC (Table 2). The results were similar when we excluded nonwhites (ADH3*1/*1: OR, 0.9; 95% CI, 0.6, 1.4; ADH3*1/*2: OR, 1.2; 95% CI = 0.8, 1.6). The distribution of ADH3 genotypes among the 29 white cases who had died before recruitment (ADH3*1/*1 = 37.0%, ADH3*1/*2 = 51.7%, and ADH3*2/*2 = 17.2%) was similar to the corresponding distribution among the interviewed, white OSCC cases (ADH3*1/*1 = 32.7%, ADH3*1/*2 = 49.0%, and ADH3*2/*2 = 18.3%). Patients with cancer of the tonsils or oropharynx had a genotype distribution (ADH3*1/*1 = 39.5%, ADH3*1/*2 = 38.2%, and ADH3*2/*2 = 22.4%) that was similar to controls, whereas patients with cancer of the tongue had a deficit of the ADH3*2 allele (ADH3*1/*1 = 38.6%, ADH3*1/*2 = 47.9%, and ADH3*2/*2 = 13.6%). The age-, sex-, and race-adjusted ORs for tongue cancer were as follows: ADH3*1/*1, OR = 1.6 (95% CI = 0.9, 2.9), and ADH*1/*2, OR = 1.7 (95% CI = 0.9, 2.9). The corresponding results for tonsillar and oropharyngeal cancers were ADH3*1/*1, OR = 1.0 (95% CI = 0.5, 1.9), and ADH3*1/*2, OR = 0.8 (95% CI = 0.4, 1.5).
The distribution of alcohol consumption (<1 drink/week, 1–14 drinks/week, 15–28 drinks/week, or ≥29 drinks/week) among controls did not differ across ADH3 genotypes: ADH3*1/*1: 26.4%, 58.4%, 8.6%, and 6.6%; ADH3*1/*2: 24.4%, 59.0%, 9.4%, and 7.2%; ADH3*2/*2: 29.1%, 53.6%, 12.7%, and 4.6%. The joint association between alcohol consumption, ADH3 genotype, and OSCC risk is shown in Table 3. ORs for OSCC increased with increasing weekly alcohol consumption for persons regardless of genotype, but the magnitude of the increase was more pronounced among ADH3*2/*2 carriers (∼10-fold between the highest and lowest alcohol consumption categories) and least pronounced among ADH3*1/*1 carriers (∼3-fold between the highest and lowest alcohol consumption categories). Combining across ADH3 genotypes, when alcohol consumption was modeled as a continuous term each alcoholic beverage consumed per week was associated with a 2.2% increase in OSCC risk (95% CI, 1.3%, 3.1%). In the extension of this model with indicator terms for ADH3 genotype and the product of ADH3 genotype and continuous alcohol consumption, the percentage increase was greatest for ADH3*2/*2 carriers (4.3%; 95% CI = 1.7%, 7.1%), intermediate for ADH3*1/*2 carriers (2.3%; 95% CI = 1.0%, 3.6%), and lowest for ADH3*1/*1 carriers (1.3%; 95% CI = 0.0%, 2.6%; Fig. 1). Comparing this model with one in which ADH3 genotype was coded continuously to represent codominance of the alleles (ADH3*2/*2 > ADH3*2/*1 > ADH3*1/*1), there was no significant lack of fit attributable to nonlinearity of the modification of the risk of OSCC associated with alcohol consumption by ADH3 genotype (P = 0.528). Comparing hierarchical models with the continuous ADH3 genotype term, the addition of the ADH3-alcohol interaction term resulted in a significant improvement in the log-likelihood (P = 0.039). The ADH3 genotype-specific percentage increases in OSCC risk associated with alcohol consumption from the model with the continuous term for ADH3 genotype were similar to those from the model in which ADH3 genotype was included as two indicator terms: ADH3*2/*2 (5.3%; 95% CI, 2.1%, 8.5%), ADH3*1/*2 (2.5%; 95% CI, 1.5%, 3.6%), and ADH3*1/*1 (1.2%; 95% CI, 0.0%, 2.4%) For intraoral cancers (n = 243), the percentage increases in risk associated with each additional alcoholic beverage were as follows: ADH3*2/*2, 3.2% (95% CI, 1.1%, 5.2%); ADH3*1/*2, 2.0% (95% CI, 0.1%, 3.1%); and ADH3*1/*1, 0.9% (95% CI, 0.0%, 2.1%). For pharyngeal cancers (n = 77), the percentage increases in risk were as follows: ADH3*2/*2, 4.4% (95% CI, 1.7%, 7.2%); ADH3*1/*2, 2.8% (95% CI, 1.5%, 4.2%); and ADH3*1/*1, 1.2% (95% CI, 0.0%, 2.8%).
The modifying effect of ADH3 genotype on the alcohol consumption-OSCC relationship seemed to be concentrated among heavy cigarette smokers. From models with continuous terms for ADH3 genotype and alcohol use, the percentage increases in risk with each additional alcoholic beverage among cases and controls with ≥20 pack-years cigarette smoking were 5.6% (95% CI, 1.6%, 9.7%) for ADH3*2/*2, 2.7% (95% CI, 1.4%, 4.0%) for ADH3*1/*2, and 1.2% (95% CI, −0.1%, 2.6%) for ADH3*1/*1; the corresponding percentage increases among cases and controls with <20 pack-years of cigarette smoking (including never-smokers) were 2.5% (95% CI, −0.1%, 8.2%) for ADH3*2/*2, 2.6% (95% CI, 0.8%, 4.4%) for ADH3*1/*2, and 2.6% (95% CI, −0.7%, 6.1%) for ADH3*1/*1.
The joint association between the ADH3 polymorphism, the GSTM1 null genotype, and OSCC is shown in Table 4. The risk was not increased among carriers of ADH3*1/*1 and GSTM1 null, either overall or among heavy alcohol consumers. Similarly, the percentage increase in OSCC risk associated with alcohol consumption did not vary consistently across the three groupings of ADH3 and GSTM1 genotypes (data not shown).
Among controls for whom dietary intake data were available (n = 360), reported fruit and vegetable consumption did not explain a significant proportion of variation in alcohol consumption beyond that accounted for by age, sex, race, and cigarette smoking. The age-, sex-, race-, and cigarette smoking-adjusted correlation between average weekly alcohol consumption and weekly dietary fruit and vegetable intake was weak (r = −0.049; P = 0.354) and did not vary consistently by ADH3 genotype (ADH3*2/*2: r = −0.016; ADH3*1/*2; r = −0.05; and ADH3*1/*1, r = −0.008).
Discussion
In this population-based study, the risk of OSCC was not associated overall with an individual’s ADH3 genotype. Rather, we found that the risk of OSCC associated with alcohol consumption was modified by ADH3 genotype. Specifically, among persons carrying two copies of the ADH3*2 allele, the risk of OSCC increased ∼5% for each alcoholic beverage consumed per week on average, compared with ∼2.5% for individuals heterozygous for the ADH3*2 allele and ∼1% for individuals homozygous for the ADH3*1 allele.
Our results seem to be opposite of those reported by Harty et al. (9) in a Puerto Rican population, who found that the strongest risk of OSCC associated with alcohol consumption was among individuals homozygous for the ADH3*1 allele. Coutelle et al. (10) observed a strong association between the ADH3*1/*1 genotype and oral cancer risk in a small study among alcoholics, a finding roughly consistent with the results of Harty et al. (9). In contrast, studies in France and the United States found no overall association with ADH3 genotypes and no evidence that these genotypes modify the effect of alcohol consumption on OSCC (11, 12, 13). With the exception of a recent study from Texas (12), our study was considerably larger than the previous investigations, and the Texas study lacked quantitative data on alcohol consumption necessary for detailed analyses of effect modification. Thus the divergent results might be explained in part by different levels of precision or data quality in estimating the relationship between ADH3 genotypes, alcohol consumption, and OSCC.
Heterogenous findings may also be caused by differences in the level and/or type of alcohol consumption among the populations studied to date. In particular, whereas ∼33% and 9% of the controls in Harty et al. (9) reported consuming 15 or more and 57 or more alcoholic drinks/week, respectively, the corresponding figures among our controls were 16% and 1.5%. Among cases and controls in Harty et al. (9) who consumed <57 drinks/week (a level of consumption more similar to that reported in the present study), the pattern of effect modification based on categorical data (see Table 3 of Ref. 9) suggests that the strongest risk associated with alcohol consumption was among the ADH3*2 homozygotes. Whereas the sparse numbers in Harty et al. (e.g., a total of 37 cases and controls who were homozygous for ADH3*2; Ref. 9) clearly limit the conclusions that can be drawn from such a comparison, the data raise the possibility that the direction and/or strength of the modifying effect of ADH3 genotype might depend on the level of alcohol consumption in the population being studied. Whether variation in the types of alcoholic beverage consumed in different populations contributes to the heterogeneity among studies is unclear, because previous investigations did not provide information on this characteristic.
Our response rates were low among both cases and controls; for cases, the low response rate was related in part to early mortality caused by OSCC. Studies of other cancers have raised the possibility that polymorphisms in metabolizing enzymes could be associated with the development of more aggressive disease or poorer survival (23, 24, 25, 26). The similarity of the ADH3 genotype distribution among recruited OSCC cases and the small number of OSCC cases we genotyped who were eligible for our study but who died before being recruited provides evidence that our findings are unlikely to be attributable to the differential survival of OSCC patients according to genotype.
Although we were able to account for cigarette smoking as the major confounder of the association between alcohol consumption and OSCC, unlike Harty et al. (9), we could not control for consumption of fruits and vegetables, which has been reported to be inversely related to the risk of oral cancer (2). We found evidence of only a weak correlation between fruit and vegetable intake and alcohol intake among the subset of controls for whom dietary intake data were available, and little consistent variation in this correlation by ADH3 genotype. On this basis, it seems reasonable to conclude that our results have not been noticeably affected by our inability to adjust for fruit and vegetable intake.
Finally, although we found that the distribution of ADH3 genotypes among white controls was statistically significantly different than what would be observed under Hardy-Weinberg equilibrium, the quantitative difference in genotype frequency was very small and unlikely to have led to spurious over- or underestimation of associations in this study. Furthermore, if ADH3 genotypes are related to alcoholism, as some studies among Caucasians have suggested (27, 28), it may be that the Hardy-Weinberg assumption of nonassortative mating should not be expected to hold for variation at this locus.
OSCC consists of lesions developing at different sites, and it is possible that associations with alcohol and/or genetic predisposing factors vary according to tumor site within the oral and pharyngeal areas. Harty et al. reported that the modifying effect of the ADH3*1 allele on the association between alcohol consumption and OSCC was present only for oral cavity cancers as opposed to pharyngeal cancer (9), whereas we found that the modification of the relationship by ADH3*2 allele was similar for both types of OSCC. Inasmuch as both studies, plus a third, that found no evidence of effect modification (11), had similar proportions of OSCC that were located in the oral cavity (range, 65–72%) as opposed to the pharyngeal area, it does not seem that heterogeneity among studies in the distribution of tumor sites accounts for the inconsistency of results. None of the studies reported to date had sufficient numbers of cases for thorough analyses of specific OSCC sites (e.g., tongue or tonsil).
It is likely that the extent to which ADH3 genotype modifies the association between alcohol consumption and OSCC could depend on other factors. Coutelle et al. found suggestive evidence of a greater than multiplicative interaction between the ADH3*1/*1 genotype, GSTM1 null genotype, and the risk of OSCC among alcoholics (10). We found no evidence that persons carrying these two genotypes were at particularly increased risk of OSCC, regardless of the alcohol consumption level. In contrast, our exploratory analyses suggested that the effect modification between ADH3*2, alcohol consumption, and OSCC was strongest among those persons with a history of heavy cigarette smoking. Such a pattern is consistent with the well-established joint effect of alcohol consumption and tobacco use on OSCC, and adds support to the validity of the observed effect modification by ADH3 genotype. Larger studies are needed to determine with greater precision the possible additional modification of the relationship between ADH3*2, alcohol consumption, and OSCC by cigarette smoking and other genetic and life-style factors.
Linkage disequilibrium between the ADH3*2 allele and alleles in other genes cannot be completely excluded as an explanation for our results or the findings from other populations. In particular, the functionally polymorphic ADH2 gene is located relatively close (∼15 kb) to ADH3 on chromosome 4q21–23, and there is a 100-fold greater difference in catalytic activity between the low- and high-activity variants of ADH2 than between the corresponding ADH3 variants (29). Studies show that 94–100% of Caucasians in Great Britain and Australia who are homozygous for ADH3*2/*2 (low activity) also are homozygous for ADH2*1/*1 (low activity; Refs. 27, 28, 30). However, because the ADH2*2 allele is carried by <5% of Caucasians of European descent, >85% of ADH3*1/*1 and ADH3*1/*2 carriers are also homozygous for ADH2*1/*1 (27, 28). If these findings apply to our population, it seems unlikely that linkage disequilibrium could explain our results for ADH3 genotypes. It is also possible that linkage disequilibrium with polymorphisms in other ADH genes (e.g., ADH4; Ref. 31) or other genes near the ADH cluster accounts for our results. Finally, as with any study in which cases are compared with unrelated controls, there exists the potential for population stratification to explain our results. Our primary finding—that ADH3 genotype modifies the association between alcohol intake and OSCC—could only be attributable to population stratification if differences in genetic background between cases and controls are related to alcohol consumption differentially by ADH3 genotype. This possibility seems unlikely, particularly because we found no evidence that ADH3 genotypes were related to alcohol consumption at the low-to-moderate levels reported in this population.
Acetaldehyde is a known animal respiratory tract carcinogen (4), and the results of two previous studies of the ADH3 polymorphism (9, 10) are consistent with a role for this compound in human OSCC. Although acetaldehyde is produced by human oral mucosa, accumulating evidence indicates that oral microflora are responsible for the vast majority of acetaldehyde production in the oral cavity, particularly among cigarette smokers and alcohol consumers, (32, 33, 34, 35). Such evidence suggests that human ADH3 genotype should have little or no influence on oral acetaldehyde production, and thus would not be associated with OSCC risk through this mechanism. Our finding that the risk of OSCC associated with alcohol consumption is modified by ADH3*2 carriership is, however, consistent with an etiological pathway that does not involve acetaldehyde. Specifically, recent studies indicate that ethanol competitively inhibits ADH3-mediated conversion of retinol to retinoic acid (36). Retinoic acid seems to have chemopreventive properties in OSCC and its precursors (37, 38, 39). If the ADH3*2 allele is directly associated with reduced conversion of retinol to retinoic acid in the same way that it is associated with reduced conversion of alcohol to acetaldehyde (a possibility which, to our knowledge, has not been studied), carriers of ADH3*2 would have, on average, less intracellular retinoic acid. This reduction would be expected to be particularly pronounced among heavy alcohol consumers, whose retinoic acid synthesis would also be impaired through competitive inhibition of ADH3 by ethanol.
Assessing the role of ADH3 polymorphisms in the development of OSCC, as well as other head and neck squamous cell carcinomas, will require additional studies that are sufficiently large to provide statistically precise estimates of effect modification across a broad range of alcohol use. Meta-analyses and pooling of data from existing studies of ADH3 genotypes investigations are methods that will help achieve increased precision, as well as investigate site-specific associations and joint associations with other OSCC risk factors (e.g., cigarette smoking). However, heterogeneity among study populations—whether in data quality, type and extent of alcohol consumption, or other characteristics—could limit considerably the success of such approaches. Thus, well-designed population-based studies with large numbers of cases and controls ultimately will be needed. These studies should include genotyping of polymorphisms in ADH2 and other ADH genes to assess more completely the contribution of genetic susceptibility to the risk of OSCC associated with alcohol consumption.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This research was supported by grants and contracts from the National Institute of Dental and Craniofacial Research (DE 12609) and the National Cancer Institute (CA 48896 and CN 05230) with additional support from the Fred Hutchinson Cancer Research Center.
The abbreviations used are: OSCC, oral squamous cell carcinoma; ADH, alcohol dehydrogenase; ALDH, aldehyde dehydrogenase; GST, glutathione S-transferase; FFQ, food-frequency questionnaire; OR, odds ratio; CI, confidence interval.
OR for OSCC as a function of average number of alcoholic drinks/week, by ADH3 genotype. Tick marks inside the horizontal axis, distribution of study participants by reported alcohol consumption.
OR for OSCC as a function of average number of alcoholic drinks/week, by ADH3 genotype. Tick marks inside the horizontal axis, distribution of study participants by reported alcohol consumption.
Characteristics of oral cancer cases and controls, Seattle-Puget Sound region, 1985–1995
Characteristic . | . | Cases (%) (n = 333) . | Controls (%) (n = 541) . | OR (95% CI) . |
---|---|---|---|---|
Age (yr) | ||||
<45 | 14.7 | 17.7 | ||
45–49 | 12.6 | 15.9 | ||
50–54 | 18.0 | 15.7 | ||
55–59 | 19.5 | 18.1 | ||
60–65 | 35.1 | 32.5 | ||
Male | 71.2 | 71.5 | ||
Race | ||||
White | 93.7 | 94.5 | ||
Black | 3.6 | 2.6 | ||
Other | 2.7 | 3.0 | ||
Less than college education | 43.8 | 28.5 | ||
Smokeless tobacco (any)a | 6.6 | 3.9 | 1.5 (0.7, 2.9) | |
Cigarette smoking (pack-yr)a | ||||
None | 15.1 | 34.8 | 1.0 | |
<20 | 14.8 | 30.9 | 1.1 (0.7, 1.8) | |
20–39 | 28.3 | 21.1 | 2.8 (1.8, 4.4) | |
40+ | 41.9 | 13.3 | 6.7 (4.1, 10.8) | |
Alcohol Use (drinks per week)a | ||||
<1 | 13.8 | 26.1 | 1.0 | |
1–7 | 25.2 | 43.1 | 1.0 (0.6, 1.5) | |
8–14 | 16.8 | 14.6 | 1.7 (1.0, 2.9) | |
15–42 | 27.0 | 12.4 | 2.8 (1.7, 4.8) | |
≥43 | 17.1 | 3.9 | 4.7 (2.4, 9.4) | |
Cigarette smoking and alcohol use combined | ||||
Smoking (pack-yr) | Alcohol (drinks/wk) | |||
Never | <1 | 7.8 | 14.8 | 1.0 |
Never | 1–14 | 5.7 | 16.6 | 0.8 (0.4, 1.5) |
Never | ≥15 | 1.5 | 3.3 | 1.2 (0.4, 3.6) |
1–20 | <1 | 2.7 | 7.6 | 0.8 (0.3, 1.8) |
1–20 | 1–14 | 8.1 | 20.7 | 0.9 (0.5, 1.6) |
1–20 | ≥15 | 3.9 | 2.6 | 3.8 (1.5, 9.4) |
≥20 | <1 | 3.0 | 3.7 | 1.8 (0.7, 4.5) |
≥20 | 1–14 | 28.3 | 20.3 | 3.3 (1.9, 5.7) |
≥20 | ≥15 | 38.9 | 10.4 | 9.9 (5.5, 17.9) |
Characteristic . | . | Cases (%) (n = 333) . | Controls (%) (n = 541) . | OR (95% CI) . |
---|---|---|---|---|
Age (yr) | ||||
<45 | 14.7 | 17.7 | ||
45–49 | 12.6 | 15.9 | ||
50–54 | 18.0 | 15.7 | ||
55–59 | 19.5 | 18.1 | ||
60–65 | 35.1 | 32.5 | ||
Male | 71.2 | 71.5 | ||
Race | ||||
White | 93.7 | 94.5 | ||
Black | 3.6 | 2.6 | ||
Other | 2.7 | 3.0 | ||
Less than college education | 43.8 | 28.5 | ||
Smokeless tobacco (any)a | 6.6 | 3.9 | 1.5 (0.7, 2.9) | |
Cigarette smoking (pack-yr)a | ||||
None | 15.1 | 34.8 | 1.0 | |
<20 | 14.8 | 30.9 | 1.1 (0.7, 1.8) | |
20–39 | 28.3 | 21.1 | 2.8 (1.8, 4.4) | |
40+ | 41.9 | 13.3 | 6.7 (4.1, 10.8) | |
Alcohol Use (drinks per week)a | ||||
<1 | 13.8 | 26.1 | 1.0 | |
1–7 | 25.2 | 43.1 | 1.0 (0.6, 1.5) | |
8–14 | 16.8 | 14.6 | 1.7 (1.0, 2.9) | |
15–42 | 27.0 | 12.4 | 2.8 (1.7, 4.8) | |
≥43 | 17.1 | 3.9 | 4.7 (2.4, 9.4) | |
Cigarette smoking and alcohol use combined | ||||
Smoking (pack-yr) | Alcohol (drinks/wk) | |||
Never | <1 | 7.8 | 14.8 | 1.0 |
Never | 1–14 | 5.7 | 16.6 | 0.8 (0.4, 1.5) |
Never | ≥15 | 1.5 | 3.3 | 1.2 (0.4, 3.6) |
1–20 | <1 | 2.7 | 7.6 | 0.8 (0.3, 1.8) |
1–20 | 1–14 | 8.1 | 20.7 | 0.9 (0.5, 1.6) |
1–20 | ≥15 | 3.9 | 2.6 | 3.8 (1.5, 9.4) |
≥20 | <1 | 3.0 | 3.7 | 1.8 (0.7, 4.5) |
≥20 | 1–14 | 28.3 | 20.3 | 3.3 (1.9, 5.7) |
≥20 | ≥15 | 38.9 | 10.4 | 9.9 (5.5, 17.9) |
All ORs are adjusted for age (continuous years), sex, and race (white, nonwhite). ORs for cigarette smoking are additionally adjusted for alcohol use (continuous average drinks/week), ORs for alcohol use are additionally adjusted for cigarette smoking (continuous pack-years), and ORs for smokeless tobacco use are additionally adjusted for cigarette smoking (continuous pack-years) and alcohol use (continuous average drinks per week).
ADH3 genotypes and allele frequencies in oral cancer cases and controls, Seattle-Puget Sound region, 1985–1995
. | Genotype frequencies (%) . | . | . | Allele frequencies (%) . | . | |||
---|---|---|---|---|---|---|---|---|
. | *1/*1 . | *1/*2 . | *2/*2 . | *1 . | *2 . | |||
Controls (n = 541) | 36.4 | 43.3 | 20.3 | 58.0 | 42.0 | |||
Males (n = 387) | 37.5 | 43.7 | 18.9 | 59.3 | 40.7 | |||
Females (n = 154) | 33.8 | 42.2 | 24.0 | 54.9 | 45.1 | |||
Cases (n = 333) | 34.5 | 47.8 | 17.7 | 58.4 | 41.6 | |||
Males (n = 237) | 35.0 | 46.4 | 18.6 | 58.2 | 41.8 | |||
Females (n = 96) | 33.3 | 51.0 | 15.6 | 58.9 | 41.1 | |||
ORa | 1.1 | 1.3 | 1.0b | |||||
(95% CI) | (0.7, 1.6) | (0.9, 1.9) |
. | Genotype frequencies (%) . | . | . | Allele frequencies (%) . | . | |||
---|---|---|---|---|---|---|---|---|
. | *1/*1 . | *1/*2 . | *2/*2 . | *1 . | *2 . | |||
Controls (n = 541) | 36.4 | 43.3 | 20.3 | 58.0 | 42.0 | |||
Males (n = 387) | 37.5 | 43.7 | 18.9 | 59.3 | 40.7 | |||
Females (n = 154) | 33.8 | 42.2 | 24.0 | 54.9 | 45.1 | |||
Cases (n = 333) | 34.5 | 47.8 | 17.7 | 58.4 | 41.6 | |||
Males (n = 237) | 35.0 | 46.4 | 18.6 | 58.2 | 41.8 | |||
Females (n = 96) | 33.3 | 51.0 | 15.6 | 58.9 | 41.1 | |||
ORa | 1.1 | 1.3 | 1.0b | |||||
(95% CI) | (0.7, 1.6) | (0.9, 1.9) |
ORs are adjusted for age (continuous years), sex, and race (white, nonwhite).
Reference group.
Joint association of ADH3 genotype, average weekly alcohol intake, and risk of oral cancer, Seattle-Puget Sound region, 1985–1995
ADH3 genotype . | OR (95% CI)a [No. cases, no. controls] . | . | . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|
. | <1 drink/wk . | 1–14 drinks/wk . | 15–28 drinks/wk . | ≥29 drinks/wk . | Totalb . | ||||
*1/*1 | 2.1 (0.8, 5.9) | 2.0 (0.8, 5.3) | 3.4 (1.1, 10.8) | 6.1 (1.9, 19.5) | 1.3 (0.8, 2.0) | ||||
[20, 52] | [52, 115] | [18, 17] | [25, 13] | [115, 197] | |||||
*1/*2 | 1.7 (0.6, 4.7) | 2.1 (0.8, 5.3) | 3.9 (1.3, 11.8) | 8.7 (2.9, 26.1) | 1.3 (0.9, 2.0) | ||||
[20, 59] | [67, 138] | [28, 22] | [44, 17] | [159, 234] | |||||
*2/*2 | 1.0 | 1.4 (0.5, 4.1) | 3.4 (1.0, 11.5) | 10.0 (2.5, 40.2) | 1.0 | ||||
[6, 32] | [21, 59] | [12, 14] | [20, 5] | [59, 110] | |||||
Totalc | 1.0 | 1.1 (0.7, 1.7) | 2.2 (1.2, 3.9) | 4.7 (2.6, 8.5) | |||||
[46, 141] | [140, 312] | [58, 53] | [89, 35] |
ADH3 genotype . | OR (95% CI)a [No. cases, no. controls] . | . | . | . | . | ||||
---|---|---|---|---|---|---|---|---|---|
. | <1 drink/wk . | 1–14 drinks/wk . | 15–28 drinks/wk . | ≥29 drinks/wk . | Totalb . | ||||
*1/*1 | 2.1 (0.8, 5.9) | 2.0 (0.8, 5.3) | 3.4 (1.1, 10.8) | 6.1 (1.9, 19.5) | 1.3 (0.8, 2.0) | ||||
[20, 52] | [52, 115] | [18, 17] | [25, 13] | [115, 197] | |||||
*1/*2 | 1.7 (0.6, 4.7) | 2.1 (0.8, 5.3) | 3.9 (1.3, 11.8) | 8.7 (2.9, 26.1) | 1.3 (0.9, 2.0) | ||||
[20, 59] | [67, 138] | [28, 22] | [44, 17] | [159, 234] | |||||
*2/*2 | 1.0 | 1.4 (0.5, 4.1) | 3.4 (1.0, 11.5) | 10.0 (2.5, 40.2) | 1.0 | ||||
[6, 32] | [21, 59] | [12, 14] | [20, 5] | [59, 110] | |||||
Totalc | 1.0 | 1.1 (0.7, 1.7) | 2.2 (1.2, 3.9) | 4.7 (2.6, 8.5) | |||||
[46, 141] | [140, 312] | [58, 53] | [89, 35] |
ORs estimated jointly within categories of ADH3 genotypes and alcohol consumption are adjusted for age (continuous years), race (white, nonwhite), and cigarette smoking (continuous peak-years).
ORs also adjusted for alcohol consumption (categories indicated in Table).
ORs also adjusted for ADH3 genotype.
ORs for oral cancer in relation to ADH3 and GSTM1genotypes, by alcohol consumption
. | ORsa and 95% CIs . | . | . | . | . | |||
---|---|---|---|---|---|---|---|---|
. | . | [No. cases, no. controls] . | . | . | . | |||
Genotypes . | <1 drinks/wk . | 1–14 drinks/wk . | 15–28 drinks/wk . | 29+ drinks/wk . | Total . | |||
[45, 141] | [137, 307] | [58, 52] | [88, 35] | [328, 535] | ||||
ADH3*2/*2 & GSTM1 not null | ||||||||
or | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | |||
ADH3*1/*2 & GSTM1 not null | ||||||||
ADH3*2/*2 & GSTM1 null | ||||||||
or | ||||||||
ADH3*1/*2 & GSTM1 null | 1.3 (0.6, 3.0) | 1.1 (0.7, 1.9) | 1.8 (0.7, 4.1) | 0.9 (0.4, 2.4) | 1.2 (0.9, 1.7) | |||
or | ||||||||
ADH3*1/*1 & GSTM1 not null | ||||||||
ADH3*1/*1 & GSTM1 null | 1.5 (0.6, 4.2) | 1.0 (0.5, 1.8) | 1.9 (0.5, 6.3) | 0.8 (0.2, 2.7) | 1.1 (0.7, 1.8) |
. | ORsa and 95% CIs . | . | . | . | . | |||
---|---|---|---|---|---|---|---|---|
. | . | [No. cases, no. controls] . | . | . | . | |||
Genotypes . | <1 drinks/wk . | 1–14 drinks/wk . | 15–28 drinks/wk . | 29+ drinks/wk . | Total . | |||
[45, 141] | [137, 307] | [58, 52] | [88, 35] | [328, 535] | ||||
ADH3*2/*2 & GSTM1 not null | ||||||||
or | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | |||
ADH3*1/*2 & GSTM1 not null | ||||||||
ADH3*2/*2 & GSTM1 null | ||||||||
or | ||||||||
ADH3*1/*2 & GSTM1 null | 1.3 (0.6, 3.0) | 1.1 (0.7, 1.9) | 1.8 (0.7, 4.1) | 0.9 (0.4, 2.4) | 1.2 (0.9, 1.7) | |||
or | ||||||||
ADH3*1/*1 & GSTM1 not null | ||||||||
ADH3*1/*1 & GSTM1 null | 1.5 (0.6, 4.2) | 1.0 (0.5, 1.8) | 1.9 (0.5, 6.3) | 0.8 (0.2, 2.7) | 1.1 (0.7, 1.8) |
ORs are adjusted for age(continuous years), sex, race (white, nonwhite), and cigarette smoking(continuous pack-years).