Tobacco Documents Online

[CANCER RESEARCH 50, 426-431, 3anuary 15, 19901 Mate Drinking, Alcohol, Tobacco, Diet, and Esophageal Cancer in Uruguay Oduardo De Stefani, Nubia Mufioz,1 Jacques Est~ve, Alberto Vasallo, Cesar G. Victora, and Sibylle Teuchmann Department of Epidemiology, Oncology Institute, Avenida Brasi1308 dep 402, Montevideo, Uruguay [E. D. S.]; lnternational Agency for Research on Cancer, 150, Cours Albert Thomas, F-69372 Lybn cedex 08, France IN. M., J.E., S. T.]; National Cancer Registry, Eduardo Acevedo No 1530, Montevideo, Uruguay [A. V.]; and Departamento de Medicina Social, Faculdade de Medicina, Universidade Federal de Pelotas, Caixa Postale 464, 96100 Pelotas, RS Brazil [C. G. ABSTRACT ~ High risk areas in South America (southern region of Brazil, . •. Uruguay, and Argentina) offer a Uniqtie opportunity to study • .Aease-eontr61. study was conducted in" Uruguay.Co investigate the role .. this problem, This is.'due to 'the ~/et that fheir pop.ulations share. ~)f mate drinking, al6ohoi, tob.aeeo, "and certain dietaryfaeto~s in the~ the habit Of drinkifig large ~luantities of a to~al tea, known by etiology of esophageal cancer. The study included 261" patients with squamous cell carcinoma of the esophagus and 522 hospital controls matched by sex and age. A strong association with a clear dose-response relationship was observed with the amount of mate drunk daily and duration of the habit. The relative risk for those drinking over 2.5 liters of mate per day was 12.2 (95% confidence intei'val, 3.8-39.6) after adjusting for the ~ffects" of age, area of resid6nce;'alcohol, and tbbacc~.~ Strong associations were also observed with tobacco smoking and alcohol drinking which appear to a~t in a multiplicative way. The relative risk for those who smoke, and drink heavily compared to that of light s~nokers and drinkers was 22.6. The risk associated with black tobacco was about three times higher than that associated with blond tobacco. A clear protective effect was found for the consumption of fruits and vegetables but a dose-response relationship was present only for fruits. Finally, an increased risk was also found for those eating barbecued meat daily. INTRODUCTION Esophageal cancer is characterized by worldwide geographi- ~e1 variation in incidence and mortality rates, even within small as (1). In South America, the highest r~ates have been re- ported in Uruguay, a small country which shows large variation in the rates of esophageal cancer. The mortality rates for males range from 40 per 10s in the northeast region, which borders Brazil, to 10 per 10~ in the capital city of Montevideo (2). The death rates are lower for females, with a male/female ratio of 3.8 for the whole country. Cancer of the esophagus can be divided into those cancers essentially due to alcohol and tobacco and those in which these two factors do not appear to play an important part. The joint effect of tobacco and alcohol exposures accounts for about 80% bf the etiology of the disease in North America (3), South America (2, 4), Europe (5), South Africa (6), and some Asian countries (7). On the other hand, in areas with extremely high incidence such as Iran and China, alcohol and tobacc6 appear to play a minor role (8, 9) and the main risk factors remain to be identi'fied..There-is e~'idence suggesting.that" opium. ~ar." may be~ the in~jo~" caose inIran (10)~ and N-'riitroso cbm'pound~'.have been proposed as possible" etiological candidate's but convincing evidence is Still lacking in China (1"1). Factors producing chronic injury to the esophagus, such as rough foods, hot beverages, and certain vitamin deficiencies, may increase susceptibility to carcinogens. Esophageal thermal injury resulting from drinking hot bev- erages is a very difficult issue to study epidemiologically because of the widespread consumption of these drinks and the unreli- of data on temperature obtaihed through interviews. Received 6/14/89; revised 9/29/89; accepted 10/16/89. 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. t To whom requests for reprints should be addressed,at International Agency for Research on Cancer, 150 Cours Albert-Thomas, 69372 Lyon, Cedex 08, France. the folk name of mate. T.hls'beverage, an infusion of the hei'b llex paraguayensis, is drunk very hot through a metal straw. Despite the fact that the prevalence rate of exposure is over 80%, there remains a well-defined nonexposed group. A pre- vious case-control study in Uruguay, in which routinely col- lected information On mate ddnkinffwas:extracted'fr0m the clinical records of patients with esophageal cancer and other cancers, showed a strong association with mate drinking (RR = 4.7, 95% CI = 1.9-12.1), after adjusting for the effects of age and alcohol and tobacco consumption (2). Although a subsequent case-control study in southern Brazil did not show a strong association between mate drinking and esophageal cancer, the moderate increase in risk and the high prevalence of mate drinking in this population could account for a considerable proportion of esophageal cancers occurring (4). This case-control study was designed to obtain further infor- mation on the hypothesis that mate ingestion is associated with the risk of developing esophageal cancer. SUBJECTS AND METHODS The basic protocol utilized in the Brazilian study (4) was used in this study. The Brazilian questionnaire was adapted to the local situation of Uruguay, especially in relation to type of tobacco and diet. In the study period from July 1985 to September 1988, 283 cases with clinical and/or radiological diagnosis of esophageal cancer were admitted to the four main hospitals in Montevideo. These centers have a catchment area which covers 45% of the population of Montevideo and about 55% of the rest of the country. Patients treated in these hospitals are covered by Social Security medical care and have rather low incomes. Conditions for eligibility were: (a) to have histological diagnosis of squamous cell carcinoma; (b) to have been diagnosed within the previous 4 months; (c) to have lived in the country for at least 5 years; (d) to be in sufficiently good physical and mental health to g!ve reliable, answers to'the questionnaire. Proxy interviews were ndt a.ccepted. Cas.es were a.scertaiaed:s.hortly after clinical diaffno.sis thr6ugh • ~he :h'o~pi~al ca.nc~r registries ~ofaerating'ih .the ~aa~. i~ipating ~ent~s. ': '" None ofth~pati~nts refused to l~e interviewe~l and 15 pakiefit~ were excluded, 12 because no histol.ogical confirmation was available and three because of a diagnosis of adenocarcinoma. Of the remaining 268 patients, seven could not be interviewed due to terminal illness. The remaining 261 patients were included in the study. For each case, two controls matched by age (_+5 years) and sex were interviewed. Condi- tions for eligibility were: (a) to be admitted to the same hospitals; (b) not to have a diagnosis o( tobacco and/or alcohol-related diseases; (c) to have lived in the country for at least 5 years. The main diagnostic categories among the controls are listed in Table 1. The questionnaires were completed by four trained social workers. Information was collected on socioeconomic status (education, income, and occupation); on the lifetime habits of drinking mate, coffee, and tea (quantity, duration, and temperature); alcohol drinking (dose, ex- pressed in milliliters of alcohol, duration, and type of alcoholic bever- age); and tobacco smoking (type of tobacco, duration, intensity, and cessation periods). Dietary habits were assessed in two time periods: 426 ICD- 550-5 360-3 574-5 153-1 600 540-.~ 174 172-1 454 185 200-7 730-" 690-~ 122 240-,' 581- 210- 281- a 14~2. the per admsis: assess 1 groups meat, p potato~ The using t unconc a mate ment i' age, an analysi tional and we age, s¢ female known urban linear adjusti contro contro yielde~ for he bined RESI At were .-.resi.dt risk | ~0 more ~ contr 03 cane, b0 Ef -,0 ers a tT, dran 03 the a ....... lifeti dura Tt very fact< the t

ESOPHAGEAL CANCER IN URUGUAY Table 1 Diagnostic categories among control patients ICD-9 Diagnosis No. 550-553 360-379 574-575 153-154 60O 540r542 174 172-173 .454 "185 200-208 186-187" 73~-7~1 690-698 122 240~)41 581-594 210-229 281-289 Inguinal and abdominal hernias Diseases of the eye Gallbladder disorders Colorectal cancer Prostatic hypertrophy Appendicitis" Breast cancer Skin cancer Varicose veins Prostatic cancer Lymphom,'-leukemia Cancer of the testis and other g~nital tumors Osteomuseular diseases .... - Other malignant tumo.rs Skin ~lisorders Hydatid cyst Disorders of the thyroid gland Urinary tract diseases Benign tumors " Blood disorders a'142; 156, 171,193. '" " . "' Table 2 Distribution of cases and controls according to main"risk factors 76 14.6 73 14.0 58 11.I 46 8.8 36 6.9 33 6.3 28 5.4 26 5.0 24 4.6 24 4.6 16 " 3.1 13 2.5 13 "2-.5 12 2.3 12 . 2.3 8 1.5 8" 1.5 7 1.3 5 1.0 4 0.7 522 100.0 Male =" Female the period just before the onset of syrripl~oms and 10 .years before admsission to the hospital. A food frequency questionnaire wa~ used to assess the consumption of the various food items. The same 10 food groups used in the Brazilian questionnaire were included, i.e., fresh meat, processed meat, barbecued meat, eggs, cereals, potatoes and sweet potatoes, vegetables, fresh fruits, fats, and dairy products, The statistical analysis was done by linear logistic modelling (12) using the unconditional method with the GLIM software (13). An unconditional analysis was done because it is generally accepted that in a matched study, this analysis can be performed if appropriate adjust- ment is made for the matching variables that in this study were sex, age, and hospital. Moreover, the results obtained with the conditional were practically identical to those obtained with the uncondi- one. Relative risk estimates were approximated by odds ratios and were calculated after adjustment either for a~e and residence or for age, sex, and residence when a common relative risk for male and female was estimated. Adjustment for residence was made because it is known that the rates for esophageal cancer are higher in rural than urban areas in Uruguay. Trend tests were performed by adjusting a linear model using the numbers of the categories as covariate and adjusting simultaneously for all relevant categorized cofactors. The control group was subdivided into controls with other cancers and controls without cancer. The statistical analysis for the main risk factors yielded similar results for both subgroups and in addition the ×2 tests for heterogeneity were nonsignificant. Thus both subgroups were com- bined in the final analysis. Case -C{~ntroi Case Confl~i Number Age <54 55-64 65-74 75+ Residence Montevideo North South Urban - Rfiral years Of study 0-2 3-4 5-6 7+ • "Incom~ (in pesos). <5,999 18.6 6,000-11,999 24.6 12,000-23,999 27.1 24,000-47,999 20.6 >48,000 9.1 199 398 62 124" 11.6a 14.6a 8.1" 5.6~ 31.1 31.9 19.3 23.4 37.7 39.2 38.7 43.6 19.6 15.6 33.9 27.4 20.1 27.1 52.8 54.'8 45..2 34.7 30.7 28.1 6.5 29.4 !7.7 24.2 20.4 41.9 25.8 50~ 40.3.. 50.0 ~8.~ 74.2 .. 68.~ 31.1 25.8 31.4 27.9 41.9 32.3 32.4 35.5 30.6 31.2 21.0 29.8 8.5 1.6 7.3 13.8 2~.8. 17.7 20.1 33.9 25.0 25.4 19.4 25.0 26.4 16.i 24.2 14.3 4.8 8.1 Smoking status Nonsmoker 3.5 14.1 66.1 84.7 Smoker 56.8 40.2 21.0 8~1 Exsmoker 39.7 45.7 12.9 7.3 Type of tobacco BIack 63.8 38.9 14.5 0.8 Blond 22.6 32.7 17.7 13.7 Mixed 13.6 28.4 67.8 85.5 Drinking habit Nondrinker 9.1 20.6 56.5 66.1 Current drinker 62.8 47.5 25.8 16.9 Occasional drinker 3.5 3.8 4.8 5.7 Exdrinker 24.6 28.1 12.9 11.3 RESULTS A total of 261 cases (199 males, 62 females) and 522 controls were included in the study. Their distribution by age, sex, resi.dence, education (years of schooling), income, and some :risk fac'tor~ is given in ~.able ~. Cases, esp.eciallF males~ ~e "more Often rural and .ha.~e Iow.er socioeconomic status, than controls. The association of these variables with esophageal cancer was evaluated as follows. Effects of Alcohol and Cigarette Consumption. Tobacco smok- ers and alcohol drinkers were defined as those who smoked or drank alcoholic beverages, for at least 1 year, independently of the amount. Cases and controls were classified according to the lifetime consumption of cigarettes and alcoholic beverages, duration and daily consumption. The distribution of alcohol and cigarette consumption being very different in males and females, the evaluation of these risk factors has been done separately for each sex. Table 3 shows Type of alcohol Wine drinker 81.5 83.6 91.2 91.7 Spirit drinker 71.7 59.7 8.3 11.1 Beer drinker 13.9 21.8 12.5 11.1 Mate status Nondrinker 2.0 9.5 1.6 8.1 Current drinker 80.4 75.4 80.7 68.5 Exdrinker 17.6 15.1 17.7 23.4 This value and subsequent values in this column represent percentages. tion. The risk increases significantly after an average consump- tion of 50 ml of pure alcohol per day in males and earlier in females. Among the latter, however, the estimates are unstable due to the small number of drinkers. The trend tests for testing • the effect 6f alcohol and tobacco are. nevertheless highly'signif- "i~.ant'(x~" wltti one" degree, of fr~ed.o'm ar~ re.si~ectively: 4,9.. and. "5.5), and the interaction between, tobacco,, alcohol, and sex-was not significarit but the ~CIs were very wide. As only two female cases and one control smoke more than seven cigarettes and drink more than 50 ml of alcohol a day, the combined effect of alcohol and tobacco could be evaluated only in males, In Table 4 the number of nonsmokers and nondrinkers were put together with those of light smokers and drinkers because the number of the former was too small (one case and 32 controls). Table 4 shows that a multiplicative model gives a good description of the data, but no formal test of this model could be done with the, limited number of cases available. The joint effect of dura- .......... ~ ..... tion of tobacco smoking and alcohol consumption yielded sire- the relative rISKS and their t~ls Ior various leVelS o!consump- . ........ ,._ ...... . nar results w!m a less gooo nt. ~ The abbreviations Used are: CI, confidence interval; RR, relative risk. " The risk for esophageal cancer decreased significantly among 427

/ ., ESOPHAGEAL CA~qCER IN URUGUAY Table 3 Relative risk for cigarette and alcohol consumpt(on adjusted four each other and for age and residence (confidence interval in parentheses) Males Females No. cases No. controls Adjusted RR No. cases No. controls Adjusted RR arettes (per day) 0 7 56 1 41 105 I-7 14 54 1.93 (0.7-5.4) 11 11 8-14 28 75 2.66 (1.1-6.8)) 15-24 62 101 4.26 (1.77~ 0.4) ~- 25+ 88 112 4.62 (1.9-11.1) J 10 8 Alcohol (ml per day) 25"-49 12 5"1 0.71 (0.3-1.6) ] • 50-149 50 117 1.37 (0.8-2.4) [ 150-249 46 38 3.57 (1.9-6.7) [ 12 250+ 49 31 5.27 (2.7-10.2) J .88 ' .. '. 24 12 I 2.~7 (0.9-6.0) 3.2~ (1.1-9.3) -.:. i ,. - i.04 (0.4-2.4) 1.89 (0.7-4.9i Table4 Age. and residetice-adjusted relative risks for the j~int effect ofalbahol ,..: number~fclgarettesper day~d~tanbtsh~own)~ " " and tb'baeco consumption Cigarettes per day . Total Alcohol (RR for (ml per day) . 0-7 8-14 15-24 25+ alcohol) 0-49 1 3.0 3.3 4.5 1 (7) (12) (16) (19) (54) 50-1492.7 2.4 6.1 6.5 1.6 (7) (5) (17) (21) (50) 150-2493.9 8.2 21.4 15.1 4.1 (3) (4) (17) (22) (46) 250-349 10.4 16.1 13.5 30.0 5.1 (I) (2) (7) (6) (16) 350+ 22.7 18.1 22.5 22.6 6.7 (3) (5) (5) (20) (33) Total (RR for cigarette) 1 1.8 2.9 3.1 (2I) (28) (62) (88) (199) exdrinkers (P = 0.01) and there was a'significant trend with years since quitting: the RR values adjusted for age, residence, and amount of cigarettes per day, with the regular drinkers as reference category, were as follows: 1-9 years since quitting 0.78 (95% CI, 0.48-1.26), >_10 years 0.46 (0.23-0.92), nondrinkers and occasional drinkers 0.48 (0.29-0.81) (P = o.ool). Influence of the Type of Alcoholic Beverage. The distribution of type of alcohol consumed by the study subjects changes with their total consumption of alcohol; as a consequence, this distribution differs between cases and controls. Light drinkers tend to drink more beer and wine and less spirits than heavy drinkers. On the other hand, heavy drinkers tend to drink more spirits than light drinkers. Since total alcohol consumption is given by the sum of the three types of alcoholic beverages,, the average linear increase for .each of them separately was calcu- late.d..adjusting for the,others and ~onsidering eaci{ I~everage..as a continudu.s vari£bl~. There Was a s!gnificant difference of effect between b~verages. Beer did not contribute to the risk, whereas wine and hard liquor increased the risk by the same amount for each milliliter of pure alcohol added. Therefore, in this population, the sum of liquor and wine consumption is a better indicator of the risk associated with alcohol consumption than total alcohol. Influence of Other Smoking Variables (Male Smokers Only). When adding to the multiplicative model including age, resi- and average cigarette and alcohol consumption, age at (four categories), number of years since quitting (five categories), and duration (five categories), no effect of age at start was found, but years since quitting smoking and duration were found to modify significantly the risk. The assessment of tobacco exposure by duration gives a better fit than the average As shown in Table 2, most of the male smokers in Uruguay smoked either black or blond tobacco. Black tobacco was used by 69% of male smoker cases against 47% among controls. In addition, 64% of cases and 40% of controls smoked black tobacco only. This difference was highly significant after con- trolling for daily dose of alcohol and tobacco. The risk for those who have smoked mainly black tobacco compared to that of those who smoked mainly blond tobacco was 2.6 (95% CI, 1.7- 3.9). For mixed smokers, mainly black tobacco smokers were those who have smoked more black tobacco than blond tobacco over their life-span. Moreover, the effect of type of tobacco combines multiplicatively with duration of smoking (Table 5). Among male smoker patients, 19% used filter cigarettes against 28% among controls. This apparent protective effect of filters was however no longer significant after adjustment for alcohol, cigarette consumption, and type of tobacco: the use of filter cigarettes was in fact similar in cases and controls after stratification by type of tobacco. No cases or controls used pipes or cigars. Dietary Factors. In the following analyses, all risk evaluations were made after adjustment for age, sex, region, alcohol, dura- tion of cigarette smoking, and type of tobacco smoked. The current and past frequencies of consumption of the 10 food groups were compared between cases and controls and no significant differences were found. Therefore the current con- sumption was used in the analysis. The 10 food groups were fresh meat, preserved meat, barbecued meat, fat, dairy products, eggs, cer.eals, potatoes, vegetables, and fresh fruits. Table 6 reports the results for those food groups which have been suspected of influencing .the risk of esop.h/~geal cancer.'A clear. • protective effect and ft. si~r~ifi~an'/ dbse-'~rsl~onse rel~ti~nshil~ • " was found with the consumption of flesh fr.uits.:A reduction risk was als.o observed with the consumption of vegetables but without significant dose response. A significant increase in risk Table 5 Age, residence, and alcohol adjusted relative risks (95% CI)a for duration of smoking and type of tobacco (males only) Type of tobacco Duration (years) Duration of adjusted for cigarette smoking type of (years) Mainly blond Mainly black tobacco~ 1-24 1 3.2 1 25-44 2.5 8.1 2.5 (1.1-5.7) 45+ 4.4 9.0 3.3 (1.5-7.3) Type of tobacco adjusted 1 2.6 (1.7-3.9) for duration a The estimation was carried out within the set of male smoker cases and controls. ~ The xz for interaction between the two factors is 1.15 with 2 df. 428 Table 6 . Ft Fresh ~ Fat Vegeta Barbec Fruits ~ Adju consumpt Dail) 0 0.( 0..~ ld 2.: 0- 15 30 45 60 ~ Adj~ tobacco. for th. withot increa: after a meat ~ group: Ma~ Only ~ was a drunk There durati the sa ~ind ifi absen~ df).Tl factor rural ' Inn nonsi. lufl ,: temp~ with" had a effect ~ mate warm signit

ESOPHAGEAL CANCER IN URUGUAY Table 6 Adjusted relative rlsks for the current consumption of.certain food stuffs DISCUSSION < once 1-3 times/ >3 times/ Trend' ":'~ ~ Food~ ' a week week week Daily (x) It is estimated that about 40% of all esophageal cancer ~s Fresh meat 1 0.30 0.38 . 0.61 2.11 -~ occurring in Uruguay during the study period were included in (0.1-0.8) (0.2-1.0) (0.2-1.5) our study. However, the degree of representativeness of the 1 1.03 2.07 1.44 2.27 (0.7-1.6) (1.2-3.5) (1.0-2.2) study cases cannot be evaluated. With this limitation in mind Vegetables I 0.49 0.48 0.56 -2.45 we can say that our results indicate that 90% of the cases of • (0.3-0.7) (0.3-0.8) (0.3-1.0) esophageal ~aneer occurred over the age of 55 and that cases Barbecued meat 1 0.86 1.04 2.66 1.73 (0.6-1.2) (0.6-1.9) (1.3-5.5) tended to live in rural areas and to have a lower socioeconomic Fruits 1 0.60 0.48 , 0.33 -4.58 status than the controls, which is in agreement with observa- (0.4L0.9) (0.3-0.8) (0.2:-0.5) tions in other populations (4i 6). With regard to etiology, this "Adjusted for age, residence, smoking duration, type'of tobacco, a~d.alcoh01 study revealed that, as in similar gtudies carried out earlier in consumption. .. ' ." -" Uruguay (2) and in Brazil (4), alcohol and tobacqo-are the main risk factbrs for this cancei- and that ~hese two factors appear Table 7 ,~djusted relative risks for mate drinking~ act in a multiplicative way. The relative risk for those who both Number Of Number oi" Adjusted RR ' ~ drink and smoke heavily (over 250 ml of alcohol/day and over cases controls (95% CI) Daily amount liters/daY ~___. 2-5 cigarettes/day) was about 20, which was less than that 0 5 48 1 reported from France (5). As regards tobacco, this population . .0.01-0.49 11 44. 2.52 offers a special opportunity to compare the effects of the two 0.50-1.49 1~3' ' : 3i7 - (0.8-8.4)3.60 " types of tobacco,/~s approximately ...... half.smoke blond, flue-cu~.ed (1.3-9.9) tobacco and the other half smoke black, air-cured tobacco• The 1.50-2.49 78 95 " 6.07. risk for smokers of black tobacco cigarettes was increased (2.1-17.3) almost threefold compared to that of smokers of blond tobacco 2.50+ 34 18 12.21 (3.8-39.6) cigarettes and it increased even more with duration than with Duration (ye~s) the number of cigarettes smoked. These findings are in agree- 0-14 7 62 1 ment with results from case-control studies on bladder cancer 15-29 11 34 3.67 in Italy (14), Argentina (15), on laryngeal cancer in southern (1.1-11.8) Europe (16) and previous studies on cancers of larynx and 30-44 58 104 4.44 (1.7-11.4) oropharynx in Uruguay (17, 18), suggesting that black tobacco 45-59 101 230 2.65 is more carcinogenic than blond tobacco. Moreover, these 0b- (1.1-6.5) 60+ 84 92 6.40 servations are reinforced by laboratory results showing that the (2.5-16.4) smoke of black tobacco cigarettes contained more aromatic ~ Adjusted for sex, age, residence, alcohol, smoking duration, and type ~5-~ amines and tobacco-specific nitrosamines than that from blond tobacco cigarettes (19) and that the urine of smokers of black tobacco contained about twice as much mutagenic activity as for those who eat barbecued meat daily was observed but did the urine of blond cigarette smokers (20). without a significant dose-response relationship. However, the In relation to alcohol, the risk increases more with the increase in risk for those who eat barbecued meat daily persisted amount of alcohol consumed than with the number of cigarettes after adjusting for meat consumption. No clear effect for fresh smoked, which is in agreement with the observations made in meat and fat and no significant associations with the other food the high risk areas for esophageal cancer in France (5). As in groups were observed, the French studies, the correlation is essentially with the Mate Drinking. Mate drinking is very common in Uruguay. amount of alcohol consumed and not with the type of alcoholic Only 9% of controls and 2% of cases were nondrinkers. There beverage. In Uruguay, the magnitude of the increase in risk • was a clear dose-effect relationship between amount of mate associated with the consumption of spirits in males was similar drunk each day and the risk of esophageal cancer (Table 7). to that associated with wine drinking, which is not surprising There was also a less convincing but significant relation with considering that about 80% of the male drinkers are wine duration of use. The slope of the dose-effect relationship was drinkers and 60% are drinkers of spirits. However, no increased the same for both sexes in each alcohol consumption category risk associated with the consumption of beer could be detected and in all smoking categories. There was, however, a significant because there are very few beer drinkers among the male drink- absence of effect amohg blond~ ~obacco smokers (xz = 8~3 on 2 ers (20%). In "contrast with the epidemiological studies which df). There was no significant interacti9n with ~ny o'tl~e~:available" " ~how that alcoh01'0.early::increases'.the. dsk..fo.resop.hagea.l. factor. The slope of the dose-effect plot was larger among the cancer even aniong .nonsmokers (21), the labo.ratory studies rural population, but not significantly so. • have yield&l negative results (22). It has long been suggeste.d Interaction terms for tobacco smoking, alcohol and mate that alcohol may act as a solvent facilitating the transport of drinking, and socioeconomic status were calculated and found carcinogens through the esophageal mucosa (23), but it may nonsignificant as expected from the size of the study, also act as a chronic irritant, raising the susceptibility to carcin- Influence of Temperature of Hot Beverages. The reported ogens by accelerating cell turnover and thus favoring contact temperature of beverages other than mate was not associated between the carcinogens and the dividing target cells (1). with the risk of esophageal cancer. The temperature of mate Concerning dietary factors, a clear protective effect and a had a nonconsistent significant effect: the slope of the dose- dose-response relationship was observed for the consumption relationship was lower among people who reported hot of fruits, which is in agreement with observations made in other mate drinking, and higher and identical among those who report populations (3, 4, 8, 24, 25). Although a protective effect of warm or very hot. The effect of dose of mate was nevertheless vegetables was also detected, no dose-response relationship was significant in the three categories of drinkers, observed. Unfortunately, no distinction was made between raw 429

and cookedvegetables. No clear effect of fresh meat was found. A protective effect was found in France (26), but an increased ~ khas been associated with fresh meat in Brazil (4). A clear rease in risk was observed for those eating barbecued food "~ily compared to those eating them less frequently, even after adjusting for meat consumption. This observation is contrary to the findings of a similar case-control study carried out in a neighboring area (4), but it is consistent with laboratory inves- • tigations showing the presence of. animal carcinogens and mu- . fagens in barbecued foods formed .by the pyrolysis of proteins (27). .. : The present study was designed to study further ~he associa- tion with mate d.rinking. A strong association with a cleardose- respon.se relationship was observed with the amount of mate drunk daily. The RR for those drinking over 2.5 liters per day was 12.2 (95% CI, 3.8-39M), 'after adjusting for age, tobacco, and alcohol. A less strong and less clear dose-response relation- ship was observed with duration of the habit. The fact that the. mate effect was present among nonsmokers and light ~mokers," and among smokers of black tobacco but not smoker9 of blond tobacco, is puzzling. The possibility that blond tobacco smskers were from a higher ~ocioeconomic level and had a higher intake I. of fruit and vegetables was considered, especially in view of a greater effect of mate drinking among the rural population, but it was not confirmed. The finding of a significant but not 2. consistent effect of the temperature at which mate is drunk is • not surprising considering the subjectivity in the perception of 3. temperature. To evaluate the degree of misclassification in the reported temperature at which mate is drunk, a validation study 4. is being carried out in Southern Brazil and in Montevideo. p~anhkere are two possible mechanisms through which mate 5. ing could increase the risk of esophageal cancer. First, the t extract may contain carcinogenic or promoting sub- 6. stances. This possibility was raised in a previous study carried out in Uruguay (2), but laboratory studies have so far not 7. demonstrated any promoting or mutagenic activity) 8. Secondly, hot mate drinking may increase the susceptibility of the esophagus to carcinogens. Several epidemiological stud- 9. ies point towards a possible effect of hot drinks on esophageal cancer incidence. Ecological studies from Japan (28), the Soviet 10. Union (29), and northern Iran (30) have suggested that inhab- itants of high risk areas drink larger quantities of hot tea than thbse of low risk areas. Also in Iran, Singapore, and Puerto I1. Rico case-control studies indicated similar differences (8, 31, 32). A prospective study carried out in Japan (25) also showed 12. a higher risk among those drinking hot green tea. A possible effect of mate drinking on precancerous lesions of I3. the esophagus has been demonstrated in an endoscopic survey, carried out in Rio Grande do Sul (33): In addition, in a recent I4~ study on chronic esophagitis among young subjects in a high risk population for esophageal cancer in China, the strongest 15. risk factor found for esophagitis was the consumption of bev- erages at burning hot temperatures.4 16. Experimental animal data suggest that hot drinks may poten- tiate the effect of esophageal carcinogens (34). The above ob- servations, and in particular the finding that mate drinking increases the risks of esophagitis, suggest that mate itself may not contain specific carcinogens but that its effect may be due 17. thermal injury of the esophagus increasing the su_s- _ a H. Yamasaki and H. Bartsch, personal communication. 18. 4 j. Chang-Claude, J. Wahrendorf, S. L. Qui, G. R. Young, N. Mufioz, M. Crespi, R. Raedsch, D. Thnrnham, and P. Correa. An epidemiologic study of chronic oesophagitis among young persons in Huixian county, Henan Province. a high-risk area for oesphageal cancer in China, submitted for publication. 19. 430 ESOPtlAGEAL CANCER IN URUGUAY ceptibility of the esophagus to carcinogens such as those con- tained in tobacco tar. Results from the previous studies in Uruguay and Brazil and the present one support this possibility. A direct test of whether mate exerts its effect through direct carcinogens or through chronic thermal injury is being carried out in Paraguay, where the habit of drinking mate is also widespread. There it is mainly drunk cold and the esophageal cancer rates are lower than in Southern Brazil, Uruguay, and northeastern Argentina. F!nally, since this study was hospital based, the possibilit~ of selection bias should be considered. The pos.gjbil!ty'of this bias. among cases cannot., be evaluated but it was attempted •~mong controls. The inclusion, of patients with other cancers i'n the control group did not appear to affect the associations detected. ACKNOWLEDGMENTS We are grateful to S. l~lacadar and L. Torr~g for their" assistan~e~ : REFERENCES Day, N. E., and Mufioz, N. Esophagus. In: D. Schottenfeld, and J. F. Fraumeni (eds.), Cancer Epidemiology and Prevention, pp. 526-623. Phila- delphia: Saunders, 1982. Vassallo, A., Correa, P., De Stefani, E., Cendan, M., Zavala, D., Chen, V., Carzoglio, J., and Deneo-Pelligrini, H~ Esophageal cancer in Uruguay. A case-control study. J. Natl. Cancer Inst., 75: 1005-1009, 1985. Yu, M. C., Garabrant, D. H., Peters, J. M., and Mack, T. M. Tobacco, alcohol, diet, occupation and carcinoma of the oesophagus. Cancer Res., 48: 3843-3848, 1988. Vietora, C. G., Mufioz, N., Day, N. E., Barcelos, L. B., Peccin, D. A., and Braga, N. Hot beverages and oesophageal cancer in Southern Brazil. A case- control study. Int. J. Cancer, 39: 710-716, 1987. Tuyns, A. J., Pequinot, G., and Jensen, O. M. Le cancer de l'oesophage en Ille et Vilaine en fonction des niveaux de consommation d'alcohol et de tabac. Des risques qui se multiplient. Bull. Cancer (Paris), 64: 45-60, i977. Segal, I., Reinach, S. G., and De Beer, M. Factors associated with oesopha- geal cancer in Soweto, South Africa. Br. J. Cancer, 58." 681-686, 1988. Jussawalla, D. J. Oesophageal cancer in India. J. Cancer Res. Clin. Oncol., 99: 29-33, 1981. Cook-Mozaffari, P. J., Azordegan, F., Day, N. E., Ressicaud, A., Sabai, C., and Aramesh, B. Oesophageal cancer studies in the Caspian littoral of Iran: results o.f a case-contrul study, Br. J. Cancer, 39: 293-309, 1979. Li, J. Y., Ershow, A. G., Chen, Z. J., Wacholder, S., Li, G. Y., Guo, W., Li, B., and Blot, W. J. A case-control study ofeosophageal and gastric cardia in Linxian. Int. J. Cancer, 43: 755-761, 1989. Ghadirian, P., Stein, G., Gorodetzky, D., Roberfroid, M., Mahon, G. A. T., Bartsch, H., and Day, N. E. Oesophageal cancer studies in the Caspian littoral of Iran: some residual results, including opium use as a risk factor. Int. J. Cancer, 35: 593-597, 1985. Lu, S. H., Montesano, R., Zhang, M. S., Feng, L., Luo, F. J., Chui, S. X., Umbenhauer, D., Saffhill, R., and Rajewsky, M. I. Relevance of N-nitrosa- mines to oesophageal cancer in China. J. Cell. Physiol., 4: 51-58, 1986. Breslow, N. E., and Day, N. E. Statistical Methods in Cancer Research. Vol. I. The Analysis of Case-Control Studies (IARC Scientific Publications, No. 32), p. 237. Lyon: International Agency for Research on Cancer, 1980. Baker, R. J., and Nelder, J. A. The GLIM System,Release 3. Generalized Linear Interactive Modelling. Oxfordz Nume~rical.Algo~thms Group, 1978. Vh~eis, P., I~steve, J., and "l~erracini, B. Bladder cancer ~n~l smoking in ~al'es: ' types of eighrettes, age at start, effect of stoppin.g and interaction with occupation. Int. 'J. Cancer, 34: 165-170, 1984. Iscovich, J., Castelletto, R., Esteve, J., Mufioz, N., Colanzi, R., Coronal, A., Deamezola, I., Tassi, V., and Arslan, A. Tobacco smoking, occupational exposure and bladder cancer in Argentina. Int. J. Cancer, 40: 734-740, 1988. Tu.yns, A. J., Esteve, J., Raymond, L., Berrino, F., Benhamou, E., Blanchet, F., Boffetta, P., Crosignani, P., del Moral, A., Lehmann, W., Merletti, F., Pequignot, G., Riboli, E., Sancho-Garnier, H., Terracini, B., Zubiri, A., and Zubiri, L. Cancer of the larynx/hypopharynx, tobacco and alcohol: IARC international case-control study in Turin and Varese (Italy), Zaragoza and Navarra (Spain), Geneva (Switzeriand) and Calvados (France). Int. J. Cancer, 41: 483-491, 1988. De Stefani, E., Correa, P., Oreggia, F., Leiva, J., Rivero, S., Fernandez, G., Deneo-Pellegrini, H., Zavala, D., and Fontham, E. Risk factors for laryngeal cancer. Cancer (Phila.), 60: 3087-3091, 1987. De Stefani, E., Correa~ P., Oreggia, F., Deneo-Pellegrini, H' Fernandez~G' Zavala, D., Carzoglio, J., Leiva, J., Fontham, E., and Riveru, S. Black tobacco, wine and mate in oropharyngeal canccr. A case-control study from Uruguay. Revue d'EpidSmiologie et de Sant6 Pubiique, 36: 389-394, 1988. Patrianakos, C., and Hoffmann, D. Chemical studies of tobacco smoke. LXIV: 3:150 20. Malay Gallet the ur blad& 21. Tuyns drinki 22. Mand Absen and e." 23. Horie II: E~ ethan~ 24, Wynd ", thees 25. Hira~ 26. Tuyns canoe. 27. Sugin found Wyn.d

LX1V: oa the analysis of aromatic amines in cigarette smoke. J. Anal. Chem., 3: 150-154, 1979. 20. Malaveille, C., Vineis, P., Esteve, J., Oshima, H., Brun, G., Hautefeuille, A., 28. Gallet, P., Ronco, G., Terracini, B, and Bartsch, H. Levels of mutagens in the urine of smokers of black and blond tobacco correlate with their risk of 29. bladder cancer. Carcinogenesis (Lond.), 10:.577-589, 1989. 21. Tuyns, A. J. Oesophageal cancer in non-smoking drinkers and in non- drinking smokers. Int. J..Cancer, 32: 443-444, 1983. 22. Mandard, A. M., Maruay, J., Helie, H., Tuyns, A. J., and Le Talaer, J.Y. 30. Absence of effect of ethanol and apple brandy on the upper digestive tract and esophagus of the Wister rat. Bull. Cancer (Paris), 68: 49-58, 1981. 23. Horie, A., Kohehi, S., and Kunatsune, M. Carcinogenesis in the oesophagus. 31. II: Experimental production of oesophageal cancer by administration of ethanolic solutign of carcinogen. Gann, 56." 429-441, 1965. 24. W3aader, E. L., and Bross, I..L A study of etiological factors in cancers of 3"2. the esophagus. Cancer (Phila.), 14: 389-413, 1961. " 25. Hirayama, T. Diet and cancer. Nutrition and Cancer, 1: 67-81, 1971. 33. 26. Tuyns, A.J.,Riboti, E.,Doornbos, G.,andPequignot, G. Dietandesophageal cancer in Calvados (France). Nutrition and Cancer, 9: 81-92, 1987. 27. Suglmura, T., Sato, S., and Takayama, S. New mutagenic heterocyclic amines 34. found in amino acid and prptein pyrolysates and in cooked food. In: E. L. Wynder, G. A. Leveille, J. H. Weisburger, and G. E. Livingston (eds.), ESOPHAGEAL CANCER IN URUGUAY Environmental Aspects of Cancer: The Role of Macro and Micro C0mpo- nents of Foods, pp. 167-186, Westport, CT: Food and Nutrition Press, 1983. Segl, M. Tea-gruel as a possible factor for cancer of the esophagus. Gann, 66: 199-202, 1975. K01ycheva, N. L Epidemiology of esophageal cancer in the USSR. In: D. Levin (ed.), Cancer Epidemiology in the USA and USSR. 3.pint USA/USSR Monograph, pp. 191-197. Washington, D.C.: NIH Publication No. 80-2044, 1980. IARC/Iran Study Group. Esophageal cancer studies in the Caspian littoral of Iran: results of l~Opulation studies. A prodrome. J. Natl. Cancer Inst., 59: 1127-1138, 1977. De 3ong, U. W., Breslow, N., Goh Ewe Hong, J., Sridharan, M., and Shanmugaratnam, K. Aetiological factors in oesophageal cancer in Singapore Chinese. Int. 3". Cancer, 13: 291-303, 1974. " Martinez, L Factors associated with cancer of the esophagus, mouth and pharyn.x in Puerto Rico.'& Natl. Cancer Inst., 42: i069-|094, 1969. Mu~oz, N., Victgra, C. G., Crespi, M., Saul, C., Braga, N. M., and Correa, P. Hot mate drinking "and precancerous lesions of the oesophagus: an endoscopic survey in southern Brazil. Int. J. Cancer, 39: 708-709, 1987. Yioris, N., Ivankovic, S., and Lehnert, T. Effect of thermal injury and oral administration of N-methyl-N'-nitro-N-nitrosoguanidine on the development of esophageal tumors in Wistar rats. Oncology (Basel), 41: 36-38, 1984. 431