Anatomical subsite discrepancy in relation to the impact of the consumption of alcohol, tobacco and betel quid on esophageal cancer

The carcinogenetic impact of risk factors on esophageal cancer (EC) may differ according to the portion of the esophagus where the tumor occurs. It is unclear why more esophageal squamous cell carcinomas (SCC) developed in the middle location. We carried out a multicenter case–control study in Taiwan to assess anatomical subsite risk discrepancy for this neoplasm in regard to the consumption of alcohol, tobacco and betel quid. Four hundred forty seven incident patients with pathology‐proven SCC of the esophagus (107 were upper‐third [U/3‐EC], 199 middle‐third [M/3‐EC] and 141 lower‐third [L/3‐EC] cases), as well as 1,022 gender, age and study hospital matched controls were analyzed by unordered polytomous logistic regression. All consumption of the three substances was related to the development of each subsite of EC, with a heterogeneously higher risk for current smokers (adjusted odds ratio (AOR) = 6.2) found in M/3‐EC and for current chewers, in U/3‐EC (AOR = 4.9). The joint risk of contracting lower two‐third EC for drinking and smoking appeared to significantly surpass those estimated by a multiplicative interaction model. Concomitant exposure to these two agents brought the risks of EC at all three subsites up to 10‐ to 23.9‐fold and additional tobacco‐free betel quid to a 30.3‐ to 75.0‐fold. In conclusion, tumor subsite discrepancy risk is related to prolonged exposure to tobacco and betel quid with inflorescence. Alcohol interacts with tobacco in a stronger supra‐multiplicative way in the middle portion of the esophagus, probably explaining why esophageal SCC occurs more commonly at this anatomical location. © 2007 Wiley‐Liss, Inc.

The carcinogenetic impact of risk factors on esophageal cancer (EC) may differ according to the portion of the esophagus where the tumor occurs. It is unclear why more esophageal squamous cell carcinomas (SCC) developed in the middle location. We carried out a multicenter case-control study in Taiwan to assess anatomical subsite risk discrepancy for this neoplasm in regard to the consumption of alcohol, tobacco and betel quid. Four hundred forty seven incident patients with pathology-proven SCC of the esophagus (107 were upper-third [U/3-EC], 199 middle-third [M/ 3-EC] and 141 lower-third [L/3-EC] cases), as well as 1,022 gender, age and study hospital matched controls were analyzed by unordered polytomous logistic regression. All consumption of the three substances was related to the development of each subsite of EC, with a heterogeneously higher risk for current smokers (adjusted odds ratio (AOR) 5 6.2) found in M/3-EC and for current chewers, in U/3-EC (AOR 5 4.9). The joint risk of contracting lower two-third EC for drinking and smoking appeared to significantly surpass those estimated by a multiplicative interaction model. Concomitant exposure to these two agents brought the risks of EC at all three subsites up to 10-to 23.9-fold and additional tobacco-free betel quid to a 30.3-to 75.0-fold. In conclusion, tumor subsite discrepancy risk is related to prolonged exposure to tobacco and betel quid with inflorescence. Alcohol interacts with tobacco in a stronger supra-multiplicative way in the middle portion of the esophagus, probably explaining why esophageal SCC occurs more commonly at this anatomical location. ' 2007 Wiley-Liss, Inc.
Key words: alcohol drinking; anatomical subsite; areca nut; esophageal neoplasms; tobacco smoking Looking at it globally, the incidence of cancer of the esophagus exhibits extensive variation. 1 Although a steep continued rise in the incidence of esophageal adenocarcinoma has been observed in Western countries during the past few decades, esophageal squamous cell carcinoma (SCC) remains the major histological type. 2 In some regions of Asia and India, more than 88% of esophageal cancer (EC) patients suffered from SCCs. [3][4][5] Studies have reported that the lower-third of the esophagus is the most common location for cancer lesion for adenocarcinoma. This is related to acid reflux, where contents from the stomach back-up into the esophagus. 2 On the other hand, SCC was diagnosed most commonly in the middlethird of the esophagus in several countries and populations. [3][4][5][6][7] It is unclear why more SCCs were observed in this subsite. Tobacco and alcohol are two principal agents involved in the etiology of EC. 8,9 Research in Japan and China has demonstrated that cigarette smoking has a stronger cancer risk in the middle-third esophagus than in the other segments. 3,4 In contrast, research in India has shown that alcohol intake affects this anatomical subsite with the highest cancer risk. 5 However, the study conducted in the US did not find an appreciable discrepancy in risk across tumor locations for these two agents. 10 In recent epidemiological studies, the chewing of betel quid, a common habit prevalent in India, Taiwan and Southeast Asia, 11 has been identified to have dissimilar risks in different anatomic sites across cancers of the oral cavity, pharynx, larynx and esophagus. [12][13][14][15][16] Discrepancies in both the quantity of the agent exerted and the duration of the mucosal contacted with the betel quid juice in target tissues might partly account for disparities in the impact of this substance on these cancers. 16 While one study, performed in India, found that a more than 5-fold cancer risk of the lower-third of the esophagus is linked to the use of this substance, 5 it is uncertain whether there is a subsite-specific effect of chewing, like those found in cancers of the upper aerodigestive tract, on this neoplasm.
Between 1995 and 2002, the incidence of carcinoma of the esophagus in Taiwan had been reported to increase by 7.2% per year, with an overwhelming majority of SCC cases (89-92%) among men. 17 Unlike some Southeast Asian countries and India, betel quid is not chewed with tobacco in Taiwan. However, betel quid consumers often concurrently have the habit of cigarette smoking or the drinking of alcohol, or both. 18 Previous studies have revealed that the use of alcohol, tobacco and betel quid is related to an elevated risk and, further, that these agents may also have combined risks produced synergistically with the genesis of EC. [12][13][14] Nevertheless, whether their carcinogenetic impact differs according to the portion of the esophagus where the tumor occurs remains unknown. The purpose of this multicenter case-control study is, therefore, to clarify the anatomical subsite discrepancies in regard to the effects of the three agents on the development of EC.

Selection of cases and controls
This is a large-scale collaborative case-control study established in three medical centers located in northern Taiwan (National Taiwan University Hospital (NTUH), Taipei) and southern Taiwan (Kaohsiung Medical University Hospital (KMUH) and Kaohsiung Veterans General Hospital (KVGH), Kaohsiung). These hospitals provide comprehensive medical services to patients of each socioeconomic situation. The detail study design for this investigation has been described previously. 13 In brief, however, we studied newly diagnosed cases of SCC of the esophagus (ICD-9, code 150) for those hospitalized in the Department of Chest Surgery and the Department of Gastroenterology at these three hospitals. Cases were identified through a system of rapid case recognition, recording selected demographic characteristics and related medical records, so that the patients could be identified and selected for our study as soon after diagnosis as possible. All suspected cancer cases were verified by histological materials, and only patients who have consistently been histologically confirmed to have esophageal SCC by the study pathologists were included.
Diagnostic images, including x-rays, MRI, CT scan and endoscopy, were extracted and used for the identification of tumor location occurring in the esophagus. Based on the criteria employed by the National Cancer Institute's SEER program, 7 cancer lesion at the anatomical subsite of the esophagus was determined according to the principle location that the tumor occurred, with distances measured from the midincisors as follows: (1)  The control subjects were recruited from the same hospitals. A system for the recognition of background data for hospitalized patients was developed and established in the Department of Preventive Medicine at each hospital. Community residents who were 1day hospitalized in these departments for their routine physical checkup at the first visit were considered eligible controls. Subjects 25 years of age or older that were confirmed to be without malignant tumors were contacted and invited to be our controls. Within 4 weeks after each esophageal carcinoma patient was identified, 1-2, or rarely, 3 eligible controls were selected and matched to each case, in terms of the study hospital, gender and age (6 3-years). A total of 1,022 controls were successfully interviewed. Of those, 918 subjects were matched to the cancer patients (447 cases) who had purely 1 anatomical subsite of cancer lesion. To obtain more robust findings, sensitivity analysis was performed and assessed in 2 series of controls. Similar results were found for the inclusion of all controls (1022 subjects) and of those (918 subjects) matched to the analyzed cases, with better precision of risk estimates in the former control group. The results from all controls analyses were presented.

Data collection
This study was reviewed and approved by the Institutional Review Boards of National Taiwan University Hospital and Kaohsiung Medical University Hospital. All participants gave written, informed permission to be interviewed and for the tracing of their medical records. Participants were interviewed by well-trained interviewers using a standardized questionnaire to draw out detailed information on demographic characteristics, substance use and daily diet.
Alcohol drinkers, tobacco smokers and betel quid chewers were defined separately as subjects who had consumed any alcoholic beverage 1 times per week, had smoked 10 cigarettes per day and had chewed 1 betel nut (measured as quid) per day for at least 6 months. Among them, current users were those who had practiced any of these habits within a 1-year period before the interview; former users were defined as those who had stopped any of the habits for at least 1 year prior to diagnoses or interviews. The information about substance use was collected with regard to the age of commencement, daily consumption, duration of the use and the type of substance consumed. For betel quid chewers, additional information was collected as to the types of substances consumed with the areca nut, which in Taiwan is usually a piece of betel leaf, stem, or inflorescence of Piper Betel Linn, or mixed. Nevertheless, not all types of betel quid products use tobacco as an ingredient. To assess the risk of the total alcohol intake, data on the types of alcoholic beverage (beer, wine, whisky, brandy, Taiwanese rice wine, Chu Yeh Ching liquor and Kaoliang spirit), the amount of ethanol consumed for each type of beverage (according to the ethanol content of each beverage and the total amount consumed), the number of drinks per day and the average frequency per week was converted to the total consumption of alcohol in grams per day. Since beer is the most popular alcoholic beverage in Taiwan, 1 drink was referred as consuming 15.75 g of alcohol, which equates to 1 regular bottle (350 ml) of beer containing 4.5% ethanol. To investigate the effect of cumulative lifetime exposure, the number of ''bottle 3 years'' and ''pack 3 years'' was calculated by multiplying the amount of the substances consumed, 15.75 g-alcohol bottles per day drunk, 20-cigarette packs per day smoked or 10-betel quid packs per day chewed, by the years of the substance used. Daily dietary habits were assessed by measuring the consumption of 20 food groups according to 3 time periods (<20, 20-40 and >40 years of age). Food intake questions were directed at the frequency and quantity for each time period. Only the consumption of the latest period for each patient was used for the data analyses.

Statistical analyses
The association between cancer at the upper, middle and lowerthird of the esophagus and the respective alcohol consumption, cigarette smoking and betel quid chewing were examined simultaneously using unordered polytomous logistic regression, which allows for statistical modeling of a dependent variable with more than 2 categories. 19 In such models, 3 sets of coefficients were separately produced by the maximum likelihood method in order to compare each tumor subsite with the controls. The odds ratio (OR), as an approximation of the relative risk, was generated by exponentiating these coefficients. To evaluate the capacity of a given risk factor to distinguish the risk of contracting EC at different subsites, the heterogeneity in the ORs between two tumor locations was assessed using the OR ratio calculated by the exponentiated difference in the corresponding two coefficients. However, only the largest difference in OR ratio was displayed. All regression models contained the matched factors, such as the study hospital (NTUH, KMUH and KVGH), gender and age (<41, 41-50, 51-60, 61-70 and >70 years old), the potential confounding variables, including the years of schooling (<7, 7-12 and >12 years of education), consumption of fruits (No, 1-7, 8-14 and >14 times per week) and of vegetables (No, 1-7, 8-14 and >14 times per week) and, where appropriate, the use of other substances as covariates. The dose-response linear trend across increasing categories of substance use was examined by assigning the median for each exposure category and treating the categories as a continuous variable. The departure from OR multiplicativity was assessed by fitting polytomous logistic regression models containing binary categorical variables, as well as their cross-products. The final model used for interpreting the summary results was assessed for the overall fit using the Hosmer-Lemeshow goodness-of-fit (GoF) test for separate pairs of logistic models. 19

Results
Cancer cases of the upper, middle and lower-third of the esophagus, as well as the controls studied, were generally comparable with regard to the hospital, gender, age and ethnicity. However, educational attainment was found to be negatively associated with each tumor subsite of EC, in that patients with higher levels of education were at lower risk (data not shown).
The distribution of three types of substance use is presented in Table I. In contrast to having the habit of betel quid chewing, an overwhelming majority of EC patients had the habit of drinking or  smoking, and this was observed in each tumor subsite. Among current-drinkers, current-smokers and current-chewers, the cumulative lifetime consumption of alcohol, tobacco and betel quid were conformably higher in the cancer cases at each portion of the esophagus than those in the controls; further, they were also found to be accordantly greater than the corresponding counterparts for the ex-users of the three substances.
The association between the aspect of alcohol consumption and the risk of EC by anatomical subsite are shown in Table II. The risks for all subsites of EC among ex-drinkers and current-drinkers were 3.5-to 4.9-fold and 5.9-to 8.5-fold higher than those among nondrinkers, respectively, with the strongest risk observed in the M/3-EC. Patients who drank with M/3-EC had a consistently greater cancer risk (adjusted OR 5 2.4-19.0) at each exposure level (years of drinking, quantity and frequency) than those who had other subsites of EC (adjusted OR 5 2.1-11.7 for U/3-EC and 2.1-12.2 for L/3-EC, respectively). However, varied subsite-specific risks across the three portions of the esophagus were not observed to be significant (OR ratios for the highest OR in M/3-EC compared with the lowest OR in L/3-EC were 0.9-to 1.6-fold, p > 0.05). A linear dose-response relationship was detected for time and intensity-related aspects of alcohol intake. All types of alcoholic beverages were associated with an increased risk of EC. Table III uses tumor subsites to display the risk of contracting EC from tobacco smoking. Current-smokers were observed to have a more than 2.5-fold elevated risk of EC than nonsmokers, with the highest risk found in the portion of the middle third. A more steep increase in risk, with respect to the length of the time  ORs and OR ratios were adjusted for age, gender, study hospital, education, consumption of vegetables and fruits, daily alcohol intake and pack 3 years of betel quid chewing.-2 Reference category. of tobacco consumed, was detected among M/3-EC patients than among cancer patients at the other locations. Further, subsite heterogeneity in cancer risk between M/3-EC and L/3-EC was linked to the exposure of tobacco smoke for >30 years (OR ratios 5 2.3-3.2). Smokers who inhaled the tobacco experienced a higher risk of EC than noninhaling smokers, and this was observed in all anatomical subsites.
The diverse subsite-specific effects of areca nut chewing on EC are presented in Table IV. Compared to nonchewers, a higher risk of EC was found among both ex-chewers and current-chewers, with significantly stronger risks (OR ratio 5 2.1-2.2) seen in the tumor portion of the upper-third (adjusted OR 5 4.6-4.9) in contrast to those in middle-third (adjusted OR 5 2.2). Cancer risks at each segment rose as the years and the amount of betel quid consumed increased. Substantially heterogeneous risks across the 3 subsites were detected among patients who chewed areca nut with a piece of Piper Betel Linn inflorescence, in that the OR ratio for U/3-EC compared with M/3-EC was 3.5-fold.
Since no cancer cases were observed in the category of nondrinking and nonsmoking chewers, the two-dimensional interaction effect of the consumption of the three substances was evaluated (Table V). Based on multiplicative interaction models, a discordant shape in the synergy effect between alcohol intake and cigarette smoking was identified across the 3 subsites of EC. The joint risk (adjusted OR 5 23.3) for drinkers who smoked appeared to significantly surpass the risk (expected OR 5 3.4) predicted from the product of the risk of each factor acting separately in M/3-EC (p 5 0.003). A similar pattern for the synergistic effect of these two agents was observed among L/3-EC patients, but not among U/3-EC patients. Alternatively, the combined effect of the simultaneous use of alcohol and betel quid, as well as of tobacco and betel quid, were well described by a noninteraction multiplicative model (p > 0.05), regardless of the subsite for this neoplasm.
Table VI exhibits the cancer risks associated with the combined use of alcohol, tobacco and betel quid. The risks were estimated from the final model that accommodates the full main effects from the three agents, the multiplicative interaction effect between drinking and smoking and the influences of possible confounders (GoF test, p > 0.05). As compared to nonusers of any type of substances, betel quid chewers had a 4.7-fold elevated risk of contracting U/3-EC, and the risk was significantly higher than that (OR 5 2.1) for M/3-EC (OR ratio 5 2.2). A combined use of alcohol and tobacco was linked to a 10-fold risk for the development of cancer lesion at each location. Further, as compared to that in L/3-EC, a heterogeneously greater joint effect from drinking and smoking was observed in M/3-EC. The greatest risk of Expected OR, predicted from the product of the risk for each factor acting separately.-2 Wald Z-tests for cross-product terms based on a multiplicative model.-3 Odds ratios were adjusted for the pack 3 year of betel quid chewing and the covariates (age, gender, study hospital, education, consumption of vegetables and fruits).-4 Odds ratios were adjusted for the pack 3 year of cigarette smoking and the covariates.-5 Odds ratios were adjusted for the daily alcohol intake and covariates.  Values in parenthises are 95% CI values. 1 Odds ratios were estimated from the model that includes the binary indicators for chewing, drinking, smoking and the cross-product of drinking and smoking, as well as the covariates of age, gender, study hospital, education, consumption of vegetables and fruits.-2 The OR for the middle-third of the esophagus was the reference group.-3 The OR for the lower-third of the esophagus was the reference group. cancer for all 3 segments of the esophagus was consistently detected among patients who practiced the 3 habits simultaneously.

Discussion
This study presents evidence that consumption of alcohol, tobacco and, to a lesser extent, betel quid, was closely related to the genesis of esophageal SCC at different anatomical subsites. While alcohol intake was found to affect the esophagus with a nondiscernible discrepancy in risk among tumor locations, tobacco smoking and betel quid chewing were observed to have a significantly stronger risk of contracting M/3-EC and U/3-EC, respectively, than those of contracting cancers at other subsites.
Alcohol abuse has been known to be involved in the pathogenesis of EC. 9 Although the independent risk of this agent on the occurrence of this neoplasm has been verified among nonsmokers, 20 there is limited information to date about its carcinogenetic impact on anatomical subsites of the esophagus. Studies conducted in Japan 3 and China 4 have reported that drinking was associated with an elevated cancer risk at all 3 segments of the esophagus, even though significant risks were confined to heavy alcohol consumption in the latter large-scale study. In contrast, findings from India have indicated that alcohol intake is responsible for only M/3-EC. 5 While a stronger cancer risk was observed in the middle esophagus at each exposure level with respect to years, amounts and frequency of drinking, our study did not identify appreciably heterogeneous risks across the three tumor locations, as shown in previous epidemiological studies. 4,10 Concerns about the location effect of tobacco use on EC have been examined in several case-control studies. [3][4][5]10 A comparatively higher risk was consistently claimed for cancers occurring in the middle portion than at the other portions in two Asian studies. 3,4 As compared to the risk for L/3-EC, our study further demonstrates that smokers had a significantly greater likelihood of producing M/3-EC as a result of tobacco exposure for at least 30 years (OR ratio 5 2.3-3.2). While reports from one Indian study showed similar findings, the risk was associated to the smoking of bidi, a common tobacco product used locally. 5 The esophagus receives most of its blood from the aorta and drains it to the liver and venal cava. The upper and lower-thirds of the esophagus are, respectively, supplied by the inferior thyroid artery and branches of the left gastric artery, and they are smaller than the branches from descending thoracic aorta, which supply to the middle-third of the esophagus. 21 The stronger risk observed in M/3-EC might be related to more abundant blood supply in this area, thus increasing the action of tobacco carcinogens in this location.
Betel quid is typically prepared using a combination of areca nut, tobacco, inflorescence, the leaf and stem of the Piper Betel Linn and slaked lime. While the major ingredients are compara-tively consistent, there is a geographic variation in the chewing of betel quid among different countries and different regions within them. 22 Alkaloid arecoline, a major component in areca nut, is its most active ingredient. Experimental evidence from in vitro studies has shown that at least 4 N-nitrosamines, two of which are carcinogens, are produced from areca alkaloids during the chewing of betel quid. 23 The Piper Betel Linn inflorescence, another important and aromatic flavored ingredient frequently added to the areca nut products in Taiwan, contains a high concentration of safrole (15.35 mg/g). 24 Genotoxic effects of betel-specific safrole on tissues from oral and EC patients have been demonstrated. 25,26 In certain areas of the Indian subcontinent and Southeast Asia, where betel quid chewing was identified as being a significant risk factor for EC, 14,22,27,28 areca nut was habitually chewed mixed with tobacco. In contrast, studies conducted in Taiwan 13 and South India 14 reported that an 2to 3-fold increased cancer risk is associated with the consumption of betel quid without the addition of tobacco. In this study, we further uncovered the heterogeneity in risk in relation to the chewing tobacco-free betel quid among different tumor subsites of the esophagus, in that a substantially higher risk for U/3-EC in contrast with M/3-EC was linked to prolonged exposure to betel quid (>20 years of chewing, OR ratio 5 2.5). Similar heterogeneous risks were noticed among the chewers who consumed areca nut with a piece of Piper Betel Linn inflorescence in it. Such chewers had the highest elevated risk (OR 5 8.0); this is lower than the significantly strong risk (OR 5 11.6-13.5) found for oral and pharyngeal cancers and greater than the nonsignificantly moderate risk (OR 5 2.9) observed for laryngeal cancer, 15,16 for developing U/3-EC in our study. As a consequence of chewing and swallowing, parts of the areca juice and like contents can flow through to the esophagus, with the upper portion being the location where the juice is first in contact with. A typical picture of betel stains in the esophagus is shown in Figure 1, which was taken from an endoscopic exam for a regular betel chewer. The discrepancy in risk with regard to carcinomas in the upper digestive tract (the oral cavity and pharynx) and upper airway (the larynx), as well as cancers in different portions of the esophagus suggests that the pathogenesis of such neoplasms might be associated with directly physical contact with betel quid juice, especially with compounds produced from the betel inflorescence.
The extent to which alcohol and tobacco interact to increase the risk of EC has been investigated epidemiologically for several decades. Most studies have shown that the joint effect of the two agents appears to be more than additive, and this is often described adequately by a simple multiplicative model 29 ; a few studies have even provided evidence of the supra-multiplicative interaction effect. 12,13 Biologically, the mechanism of alcohol (perhaps acting as a solvent) making physical contact with tobacco carcinogens, thereby facilitating the entry of the carcinogens into epithelia cells, is one possible explanation for such an effect. 30 Another likely reason for this is that ethanol inhibits hepatic metabolism of tobacco carcinogens, thereby increasing the internal dose of carcinogens in the esophagus. 31 In our study, a significantly multiplicative interaction between the two substances was found in esophageal carcinoma in the lower 2/3 portions, with a higher effect noted in M/3-EC than in L/3-EC (OR ratio for the joint risk 5 2.4, p < 0.05). As discussed earlier, the stronger interaction effect between the two agents in M/3-EC might be related to more abundant blood supply in this region, thus strengthening the joint carcinogenetic action of alcohol and tobacco at this location. Since M/3-EC had the largest proportion of patients who concomitantly practiced the habits of drinking and smoking (79.9%), the stronger supra-multiplicative synergistic effect detected at this location might account for why SCC occurs more commonly at this subsite.
The degree that the chewing of areca nut modifies the effects of alcohol intake and tobacco smoking with regard to the occurrence of EC has been evaluated in several locales where chewing is prevalent. In South India, a multicenter case-control study reported that chewing (treating chewers with or without the addition of tobacco as a category) significantly modified both the risk of drinking and smoking in a multiplicative way. 14 In Taiwan, only a supra-additive interaction for these relationships was found. 13 One study, performed in regions with an high incidence rate of EC in India, proposed that the joint risk of M/3-EC due to bidi smoking and nontobacco chewing was greater than the sum of the risk for each factor acting alone. 5 Our research results demonstrate that the combined carcinogenic effect of chewing and drinking, as well as of chewing and smoking, was well predicted by a multiplicative model in all subsites, with the highest joint risks being seen in U/3-EC, the segment of the esophagus nearest the oral cavity.
As compared to the abstainers, our study shows that concomitant exposure to alcohol and tobacco brought the risks of EC at all three subsites up to 10-to 23.9-fold, and additional betel quid to 30.3-to 75.0-fold (Table VI). This apparent risk excess due to chewing suggests the presence of a synergistic effect from tobacco-free betel quid on the joint risk of drinking and smoking, regardless of the subsites for this neoplasm. The 3-agents combined risk for U/3-EC (OR 5 75.0) was found to be lower than those for cancers of the oral cavity (OR 5 122.8) and pharynx (OR 5 96.6), but higher than that for cancer of the larynx (OR 5 40.3) in studies conducted in Taiwan. 15,16 The synergistically carcinogenetic impact as a result of betel quid chewing might be linked to the anatomical site in the upper aerodigestive tract.
In this report, a high degree of consistency in regard to the major associations explored across educational levels provided some assurance that these findings reflect real differences. Under-reporting of smoking and drinking among control subjects here was unlikely since both proportions for the users of the two substances in our study were analogous to those reported in two large-scale studies designed using the methods of a cross-sectional and cohort study. 32,33 While the proportion of male chewers herein was somewhat lower than those estimated from the latter two studies, this should not substantially alter the major results observed for the chewing of betel quid.
In summary, clearly the consumption of alcohol, tobacco and betel quid are involved in the etiology of EC. Prolonged exposure to tobacco and betel quid with inflorescence accounts for part of the discrepancy in risk across tumor locations. Furthermore, alcohol interacts with tobacco in a stronger supra-multiplicative way in the middle portion of the esophagus, probably explaining why esophageal SCC occurs more commonly at this anatomical location.