Document zQnD9EX2nak3RaJ8oe59pXw4R

(CANCER RESEARCH 48, 3843-3848, July I, 1988) Tobacco, Alcohol, Diet, Occupation, and Carcinoma of the Esophagus1 Mimi C. Vu,2 David H. Garabrant, John M. Peters, and Thomas M. Mack Department of Preventive Medicine, University of Southern California School of Medicine, Los Angeles, California 90033 ABSTRACT Information on occupation, smoking, food and beverage consumption, and medical history were compared between 275 incident cases of carci noma of the esophagus and 275 neighborhood controls who were matched to the cases on age (within 5 years), race, and sex. Tobacco use, mainly cigarette smoking, was a significant risk factor for carcinoma of the esophagus. Ex-smokers of cigarettes showed a reduced risk relative to those who continued to smoke, and current smokers of two or more packs per day displayed a higher risk than those who smoked less. Alcohol consumption was another significant risk factor for carcinoma of the esophagus; there was a highly significant trend with average daily dose of ethanol. Relative to controls, cases also consumed significantly more fried bacon or ham, less fresh fruits and raw vegetables, and were more likely to prefer white than whole grain bread. Finally, there was a significant association between carcinoma of the esophagus and long- term occupational exposure to metal dust; this association was largely confined to the lower one-third section of the esophagus. 20-64 years. Cases were identified through the Los Angeles County Cancer Surveillance Program (5), a population-based cancer registry which records all cases of cancer that are microscopically verified or mentioned on a death certificate. We attempted to recruit all cases diagnosed between January l, 197S and March 31, 1981. For all potential cases, we sought permission from the attending physician prior to contacting the patient or family for an interview. We began interviewing cases in April 1975 and terminated case recruitment in August 1981. Controls were selected from the neighborhood of the cases' residence at the time of diagnosis. Using the house of each case as a reference point and proceeding in a systematic and invariable sequence, we canvassed up to 80 residential units to identify a control matched to the case on sex, year of birth (within 5 years), and race (non-Hispanic white, Hispanic white, black, Asian). We attempted to identify the sex, age, and race of the residents of each housing unit. A letter was left at each unattended unit to complete the census; two additional letters were sent at 3-week intervals if the previous one was unanswered. Our INTRODUCTION The incidence of esophageal carcinoma shows remarkable variation in its geographical distribution (1). In the high risk areas of Central Asia, dietary deficiencies are believed to play a major role in the pathogenesis of this disease (2). In the United States, carcinoma of the esophagus is relatively uncommon, although there is a severalfold difference in incidence between the black and the white population (3). In Los Angeles, the average annual age-adjusted incidence of esophageal cancer during 1972-1982 in black men, black women, white men, and white women was respectively 16.4,4.9,4.1 and 1.7 per 100,000 person-years. There is some evidence that the differential in risk of esophageal cancer between United States blacks and whites has continued to widen; in 1976-1980, the mortality rate in black males under age 55 was more than six times that in white males of similar ages (4). Consumption of tobacco and alcohol have been repeatedly shown to be major causes of carcinoma of the esophagus; it has been estimated that up to 80% of cases in industrialized countries could be attributed to exposure to these two environmental factors (1). However, it is not clear how much of the difference between United States blacks and whites can be explained by variations in exposure to tobacco and alcohol, and there is speculation that dietary defi ciencies or occupational exposure might be responsible for some esophageal cancer. In this report, we describe a case-control study that was designed to study blacks and whites, to select representative cases and controls, and to search for previously unrecognized causes and the explanation for the observed racial difference. MATERIALS AND METHODS We studied histologically confirmed incident cases of carcinoma of the esophagus occurring among residents of Los Angeles County aged Received 12/18/87; revised 4/4/88; accepted 4/5/88. 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. 1Supported by USPHS Grants CAI9496, CAI7054, and CA00884 from the Division of Extramural Activities, National Cancer Institute. Presented at the Fifth Symposium on Epidemiology and Cancer Registries in the Pacific Basin held in Kauai, Hawaii, November 16-21, 1986. 2To whom requests for reprints should be addressed. goal was to interview the first resident in each sequence who met the matching criteria. If the individual refused, the second eligible control in the sequence was asked to participate. If no potential control was identified within 80 housing units, then the race-matching criterion was relaxed. We excluded cases for whom no age- and sex-matched controls could be secured. At the close of case recruitment, 488 eligible patients were identified by the cancer registry. We were unable to locate 77 cases, the physician refused to cooperate in 33 cases, the patient or family refused to cooperate in 87 cases, there was a language barrier in 8 cases, and we failed to secure a matched neighborhood control for 8 cases. Therefore, 275 (56% of eligible patients) pairs of cases and controls were included in the study. The matching criterion on race was not met for 25 controls. Most of the controls (239) were the first ( 161) or the second (78) eligible neighbors. Almost all interviews were conducted in the homes of the respondents (98% of case interviews and 96% of control interviews). The remaining interviews were conducted by telephone at the request of the respond ents. One hundred twenty-nine (47%) cases were interviewed directly; due to death of the patients, the remaining case interviews were con ducted by proxy with close family members. Fifty-nine % of these proxy interviews were conducted with the spouse, and 37% were with a first degree relative. At the start of the study, we attempted to obtain a proxy interview from the analogous member of the control family if the index case was interviewed by proxy. This proved to be extremely difficult, so of the 146 controls matched to cases interviewed by proxy, only 55 could also be interviewed by proxy; the remainder were inter viewed directly. The information from paired direct interviews was therefore analyzed separately in parallel to that from all pair interviews. All interviews were administered by a single interviewer using a structured questionnaire. The interview took approximately l h and covered lifetime occupational history, vocational and avocational ex posures to specific substances and industrial processes, smoking, alco hol and beverage history, usual frequency of consumption of a few broad food groups (beef, fried bacon or ham, deep fried foods, milk, eggs, barbecued or smoked meat, smoked fish, fresh fruits or raw vegetables, whole grain bread, and various ethnic categories of food), prior medical conditions, family history of certain diseases, and use of selected drugs. For cases, all events occurring after the diagnosis of cancer were eliminated. Similarly, events first occurring in the life of each control after the diagnosis of the index case were excluded. Subjects were asked to recall specific workplace exposures, but such recollections are likely to be unreliable for those interviewed by proxy. Therefore, when we noted an association with reported metal dust exposure among directly interviewed pairs of cases and controls, we attempted to confirm the association in all cases and controls by using 3843 TOBACCO, ALCOHOL, DIET, OCCUPATION, AND ESOPHAGEAL CARCINOMA an exposure index which was derived from the titles of jobs previously held for a minimum period of 6 months. The assumption was that one is more likely to obtain accurate information regarding job titles than specific exposures from next of kin. Formulation of a matrix by which job titles are linked to categories of dust exposure has been described previously (6). Briefly, we reviewed occupational titles from the 1970 United States Bureau of Census Index and assigned each to 1 of 10 exposure categories: no exposure to particulates, exposed to metal dust only, mineral dust only, organic dust only, metal and mineral dusts, metal and organic dusts, mineral and organic dusts, smoke and exhaust, generally dusty jobs, and unknown exposure. Each subject was classified as exposed or unexposed according to whether the individual had ever held a job title assigned to that category. Duration of exposure was calculated by adding up the years worked in those jobs. We used standard matched-pair methods (7) to analyze questionnaire data from the 275 pairs of cases and matched controls. Study variables were examined individually and then simultaneously for confounding and interaction effects. We used the exact binomial test on individual dichotomous variables. The multivariate conditional logistic regression method was used on single variables with more than two possible outcomes as well as for multivariate analysis. Pairs in which either the case or the control failed to answer the relevant question were elimi nated from the analysis. All statistical significance levels (/' values) quoted are two-sided. The method of Bruzzi et al. (8) was used to compute the population attributable risk. When we computed total ethanol intake from reported amounts of beer, sweet wine, table wine, and spirits, we used the following conversion factors: 1 can (12 fl oz) of beer contains 12.96 g; l glass (3.5 fi oz) of dinner wine contains 10.10 g; 1 glass (3.5 fl oz) of sweet wine contains 15.76 g; and 1 jigger (1.5 fl oz) of spirits contains 14.03 g of ethanol (9). 129)Never Table 1 Use of tobacco among cases and controls Directly interviewed pairs (N= 275)Cap/aCiors18(/N73=11/2221/3111/1810/764/5683 CL0.7,1.5,1.3,1.7,3.1,7.7,4.5,1.6,1.5,1.1,1.7,1C.0L,50.059,0.6.61,10.82,04..62,126.3.5,64.6.71.291.28.1812.87.2601.91.83.22.197, .112.3" used tobacco CigarettesEx-smokerQuit before2Q0u+ityr before1Q0u-1it9 yr before5Q-9uityr beforeCu0r-r4enyt r smoker1 lespsa2ck/day or packs/day3+ 29.01.5,1.4,0.2,0.3,1.1,0.4,16.917.58.95.822.411.7All packs/dayCigars1-4/dayS+/dayPipe1-4/day5+/dayCigars and/or pipe onlyCa/Co"11/317/119/199/103/339/2632/1414/810/82/55/146/54/6RR1.01.92.03 Ca/Co, cases/controls; CL, confidence limits. Table 2 Consumption of alcoholic beverages among cases and controls Directly interviewed 129)NondrinkerBeerWeekly1 pairs (A' = =275)Cap/Caiors8/3(A96' 0/46106/7228/1445/7120/1031 CL2.5,2.1,1.8,0.9,2.0,2.9,1.3,0C.9L,02..88,,13..05,,24..11,,31..71,,04..91,,11.25.,52,.10.,62,. ca-6ns/day7+ cans/daySweet wineWeekly1-3 RESULTS Sixty-eight % of the cases were men, 66% were non-Hispanic glasses/day4+ glasses/dayDinner wineWeekly1-3 whites, and 88% were aged 50-64 years. Of the 87% cases with glasses/day4+ information on tumor location in the esophagus, the up- glasses/daySpiritsWeekly1 per:middle:lower one-third ratio was 1:1.8:2.1. The predomi nant (88%) histolgica! diagnosis was squamous cell carcinoma; the remaining cases were either adenocarcinomas (8%) or "car cinoma, unspecified" (4%). A comparison of the 275 cases studied with the 488 eligible cases showed the studied cases to sh-o3ts/day4+ shots/dayAverage ofethdaaniolyl intake (g)less 4040t-h7a9n80-119120+Ca/Co"5/2032/1949/3913/929/379/39/245/4612/132/146/6436/2538/1641/5534/28 be representative of all eligible patients with respect to sex, race, age, religion, place of birth, marital status, social class, subsite, and histology. " Ca/Co, cases/controls; CL, confidence limits. 34.95.9,41.1 Cases and controls were similar in their age distribution. The mean age at diagnosis of the cases was 56.5 years, while that of esophageal carcinoma. This elevation in risk remained when the controls at date of diagnosis of the index case was 56.3 we restricted exposure to cigars and/or pipe in the absence of years. Cases and controls were comparable in their distributions cigarettes (Table 1). by religion and by marital status. However, cases had signifi cantly lower levels of education compared to controls (P for linear trend = 0.02). Whereas 17% of controls graduated from college, only 11% of cases did so. Cigarette smoking was a strong predictor of risk for carci noma of the esophagus (Table 1). Individuals who had ever smoked cigarettes regularly exhibited a significantly increased risk for carcinoma of the esophagus relative to those who had Consumption of alcoholic beverages was another strong pre dictor of risk for carcinoma of the esophagus (Table 2). Relative to individuals who did not drink regularly, regular drinkers (those who drank at least once a week) of beer, sweet wine, dinner wine, and spirits all showed a highly significant dose- response relationship with usual quantities of consumption. Increased risks were observed when exposures were limited to beer only, sweet wine only, and spirits only (there were no cases never used any tobacco products on a regular basis. Relative to who drank dinner wine only). The multivariate conditional those who continued to smoke, ex-smokers showed a reduced logistic regression method was used to examine the effect of risk for carcinoma of the esophagus, and the magnitude of the reduction in risk increased with the length of time since cessa tion of smoking. Among current smokers, those smoking one pack or less per day had a lower risk for carcinoma of the esophagus relative to heavier smokers. There was no further increase in risk for esophageal carcinoma in current smokers smoking 3 or more packs per day relative to those smoking individual beverages (each converted to g of ethanol) in the presence of all other alcoholic beverages. Consumption of beer, sweet wine, and spirits was found to exert significant independ ent effects on risk for carcinoma of the esophagus. Of those three types of alcoholic beverages, spirits showed the strongest and beer the weakest association with carcinoma of the esoph agus when the amount of ethanol was held constant. For all about 2 packs per day. Relative to nonusers of tobacco products, pairs, the regression coefficients for spirits, sweet wine, and pipe and cigar smokers both had significantly elevated risks of beer per 10 g increase in ethanol content were 0.173, 0.147, 3844 TOBACCO, ALCOHOL, DIET, OCCUPATION, AND ESOPHAGEAL CARCINOMA Table 3 Selected food consumption among cases and controls of cancer; the mean interval was 30.8 years before diagnosis. Directly interviewed pairs (N = 129) All pairs (N = 275) Nine cases had been exposed for 15 or more years; the mean duration for all 12 cases was 19.8 years. Ten cases (and both CL0.6,1.0,0.9,1.1,0.C6,L00.7.3,0,0.6.8,0,2R.5.R9,02..096/,012.1621,16.95/,115.30,718.6/9,26.17323.3/125.1939c.15o/18n.18t41r1o.6/l21s5.)5913./3a937l.s30o84/.34r73e1.2p338o./4r13t3e.57d5C8L/5,e4x1p3o9/s1u3r2e71/8to2 other kinds of metal dust Beef1/wk less2-4/wokr5+/wkFried on those same jobs which exposed them to beryllium. There was no association between reported occupational exposure to hamb1a/wcokn or less2-4/wokr5+/wkMilk1/wk asbestos and esophageal carcinoma risk (18 cases and 25 con trols, RR = 0.6, P = 0.28). 2.01.4,0.7,0.8,0.5,0.6,0.4,0.9,1.4,1.1,1.3,8.23.0T2a.4b2l.e02.432.s7h6.o6w4.5s3.7th4.e3Cda/iCffoerence between cases and controls in their prevalence of exposure to various kinds of dust. Exposure less2-4/wokr5+/wkEggs1/wk information was derived from job titles ever held by the case or the control for 6 months or longer before diagnosis of the case (see "Materials and Methods" for a more detailed description). less2-4/wokr5+/wkBarbecued Occupational exposure to metal dust was a significant risk factor for carcinoma of the esophagus. There was no association meat1/wkor smoked less2+/wokrFresh (237/24931/19.3.5.0.0.4.or.0.1.1w.0i.0th).8.o0.t7h3e3r/20kinds of dusts. Similar results were obtained when jobs that started within 10 years of diagnosis of the case were excluded. fruits or raw vegeta There was a duration-response relationship between exposure bles1/wk less2-4/wokr5+/wkBread 2.3'101/63 to metal dust and risk for carcinoma of the esophagus. The preferenceWhole grainNo 2.2133/184 1.067/113 1.082/67 results were similar whether or not we excluded jobs that were started shortly prior to diagnosis of the case. Table 5 presents the results based on exclusion of jobs started within the prior preferenceWhite* 2.1121/90 2.4a95% 10 years. In all instances, higher risk was associated with longer caCseas/C/co,ntrols;* 0.L05in,cear trend P < limitsc.obnuftidence 0.01.>but duration of exposure. The association was also subsite specific, the increased risk was largely confined to the lower one-third 0.L0i1n,edar Linear trend trend P P < < 0.001Ca/C>o"08./0101179/6742/5167/7937/4124/839/5019/1771/6226/285s8e/6cT1t4iho4e/n39a1ob2fo0/vt1he2e09ri/e9ssk1o3p/f8ha5ac1g/t2ou7rs6s2/9w(1Te3a5rbe/6l0ef3u85r/t2)h9.e56r/40eRxRa.m0.5in.0e.0d.13w.4ith1i.n01.4s1p.e4c1.i0f1ic.01.21.01.02.5'2.51 and 0.079, respectively. For directly interviewed pairs, the corresponding regression coefficients were 0.183, 0.124, and 0.025, respectively. When all sources of exposure to alcohol were summed to form a single index (average daily intake of ethanol), a highly significant trend was observed between the alcohol index and carcinoma of the esophagus (all pairs and directly interviewed pairs, P < 0.0001). Table 3 presents the distribution of consumption frequency of various food groups in cases and controls. Subjects were asked to choose between three specified categories of frequency: sex and race groups (white males, white females, black males, black females). Albeit the small numbers in the subgroups, there was remarkable consistency in the strength of the associ ations among the four groups. We also examined the risk factors within the 161 case-control pairs in which the first eligible control participated; similar results were obtained. We used the multivariate conditional logistic regression method to examine the joint effect of the various risk factors (Table 6). Each factor remained significant after simultaneous adjustment for the other factors and the highest level of education. Aside from the risk factors reported above, cases were signif- once a week or less, two to four times a week, and five or more times a week. Cases and controls were similar in their frequency of consumption of beef, milk, and eggs. However, cases had Table 4 Occupational exposure to dust among cases and controls (exposure information derived from job titles) significantly higher frequency of consumption of fried bacon or ham, and significantly lower frequency of consumption of fresh Directly interviewed pairs (N= 129) All pairs (N= 275) fruits or raw vegetables relative to controls. Also, significantly CL0.7, CL0.7, more cases than controls preferred white over whole grain bread. A multivariate analysis of these three dietary factors showed a significant association between each factor and risk for carcinoma of the esophagus after adjusting for the other two. We also examined the effects of smoking, drinking, and diet by tumor location (upper, middle, and lower third of the esophagus). No appreciable differences were observed. Any dust Metal 2.7 1.4, 5.4 84/63 1.6 1.0,02.6.5, Mineral 26/24 1.1 0.6, 2.3 51/50 1.0 1.7 Organic 35/42 0.8 0.5, 1.3 63/82 0.7 0.4, 1.0 Smoke and exhaust 15/24 0.6 0.3, 1.2 39/43 0.9 0.5, 1.4 Exposed, generally 61/52RR1.13.495%0.8,22.5.3Ca/C12o51/8191/61R84R11..1295%0.8,1.17.7 dusty jobsCa/Co93/8847/27 C a/Co, cases/controls; CL, confidence limits. When we compared the prevalence of reported exposure to metal dust on the job between directly interviewed cases and Table 5 Duration of metal dust exposure started 10+ years before diagnosis of the case (exposure information derived from job titles) controls, we observed a significant excess among cases (42 cases and 29 controls, RR3 = 2.2, 95% confidence limits =1.1, 4.5). Directly interviewed pairs OV= 129) All pairs (N = 275) Exposure to every kind of dust (beryllium, chromium and chromtes, nickel, other metal dusts) was more prevalent among cases than controls. The association was particularly strong for exposure to beryllium. Twelve cases and only two controls claimed a history of beryllium exposure on the job (RR = 6.0, 95% confidence limits = 1.4, 37.5). Initial exposures of these 12 cases all occurred at least 19 years before the diagnosis 1The abbreviation used is: RR, relative risk. Ca/Co" RR 95% CL Ca/Co RR 95% CL Nexoposure1-19 yr20+ yrSubsite: lowerone-thirdNo 12.80.5, 1.71.9, 11.30.4, exposure1-19 yr20-1- 5.81.2, 2.01.7,20.1 yr83/10327/2019/631/406/613/41.02.041.75.16.10918.6/241.6514.60/,540.0218./67,77/8711/1619/41.01.14.61.00.95.80.7 ' Ca/Co, cases/controls; CL, confidence limits. 3845 TOBACCO, ALCOHOL, DIET, OCCUPATION, AND ESOPHAGEAL CARCINOMA Table 6 Adjusted regression coefficients for various risk factors of carcinoma of the esophagus Directly interviewed pairs (N= 129) All pairs (N = 275) relative to current smokers, and current heavy smokers dis played a higher risk than current light smokers. There was no further increase in risk with daily dose of cigarettes beyond two B"Average Adjusted P0.0320.0050.0B1"700..00235300..02142000P..4089.09600A0.4d01j.u09s00t0.e30d3.0900p8.a00c4.0k32Ts4w.0o.0A-s2il4dc0eo.d0h1o7l consumption, on the other hand, exhibited a monotonie increase in risk within the reported range of daily oftobacdcaoily use (g)Average ofethandoalily intake consumption. We estimate that 93% of our cases were associ ated with exposure to tobacco and alcohol either singly or (hga)mFr*iFebdreaschon or rawvegfertuaibtsleso*rBread preference'Duration dustexpoosfurmeetal 10+yr started jointly. We observed differential risks for carcinoma of the esophagus by different types of alcoholic drinks. For any given level of ethanol intake, the risk from spirits was more than two times that from beer, with risk from wine intermediate between those ofcbaseefo(yrer)0d.i0a1g8n0o.s0i0s80.5620.5820.2150.033Two-sided from spirits and beer. Our findings support those of other " Parameters estimated from logistic regression model which included highest level of education and all factors in the table. * Coded levels are: 1/wk or less = 1; 2-4/wk = 2; 5+/wk = 3. ' Coded levels are: whole grain = 1; no preference = 2; white = 3. studies. In Japan (21), the excess risk associated with alcohol was highest for whiskey and shochu (a local liqueur) and lowest for beer. In Washington, DC (23), the excess risk was greatest for hard liquor, particularly whiskey or bourbon, and lowest for icantly more likely than controls to report a history of stomach ulcer, frequent sore throat (more than 2 times per year), chronic cough, and regular use of antibiotics. Frequency of use of table beer. In New York, Wynder and Bross (22) reported a higher risk for heavy whiskey drinkers than for heavy beer drinkers among those smoking between 16 and 34 cigarettes per day. In Puerto Rico (24), there was little excess risk among those salt and condiment was significantly higher in cases than in controls. The relative weights (10 years before) of cases were significantly lower than those of controls. Among females, cases achieved menopause significantly earlier and were significantly less likely to have hormone replacement therapy than controls. None of these associations remained significant after adjust ment for highest level of education, tobacco use, consumption drinking solely beer; among drinkers of spirits, the greatest risk was in those who took drinks straight. Both descriptive and analytical studies have implicated die tary deficiencies as a risk factor for carcinoma of the esophagus. High-risk populations have been noted to be deficient in vita mins, trace metals, and essential amino acids. Case-control studies in diverse populations have reported significantly lower of alcohol, bacon or ham, and fresh fruits or raw vegetables; intake of food groups high in micronutrients among cases preference for white versus whole grain bread, and occupational relative to controls (1). In the United States, Wynder and Bross exposure to metal dust. There was no association between carcinoma of the esopha gus and frequency of tea consumption. The reported tempera ture of the tea and the manner it was drunk (sipped slowly, gulped quickly, neither) were not different between cases and controls. Similarly, the frequency, temperature, type (caffein- ated, decaffeinated, both) and manner of consumption of coffee was not related to carcinoma of the esophagus. DISCUSSION The role of tobacco and alcohol use in the etiology of carci noma of the esophagus is well established. Elevated risk in smokers relative to nonsmokers has been consistently demon strated in cohorts of United States Veterans (10,11), American Cancer Society volunteers (12), British physicians (13), mem bers of labor unions (14), and representative residents of Japan (IS). As a group, alcoholics suffer a high risk (16-18), just as religious nondrinkers enjoy a low risk (19, 20); moreover, a significant independent effect of alcohol consumption was ob served in the Japanese prospective cohort (21). Case-control studies conducted in New York (22), Washington, DC (23), Puerto Rico (24), Uruguay (25), Brazil (26), France (7, 27), and Singapore (28) have provided further evidence of the close and independent roles of tobacco and alcohol in the etiology of carcinoma of the esophagus. In contrast to most earlier case-control studies which are (22) reported that mostly white cases in New York consumed significantly less milk, and fewer green and yellow vegetables than controls. In a study of black men in the Washington, DC area, Pottern et al. (23) observed 2-fold increased risks among individuals in the lowest, compared to the highest, tortile of consumption of each of three food groups; fresh or frozen meat and fish, dairy products and eggs, and fruits and vegetables. Recently, two studies in Europe (29, 30) showed significant negative associations between consumption of fresh meats, vegetables, and fruits and esophageal cancer risk. Our cases consumed significantly fewer fresh fruits and raw vegetables relative to controls, in agreement with the results of previous studies. On the other hand, we did not find any difference in consumption of beef, milk, or eggs between cases and controls. We observed a decreased risk of esophageal carcinoma among individuals who preferred whole grain bread, but this is likely to be a surrogate measure of better nutritional status rather than a direct measure of a specific nutrient defi ciency. The crudeness of our dietary questions preclude any examination of risk of esophageal carcinoma by level of dietary micronutrients. We demonstrated a significant dose-response relationship between intake frequency of fried bacon or ham and risk of esophageal carcinoma. These cured meat products contain ni trosa mines (31), many of which are potent inducers of malig nant tumors, including esophageal carcinomas, in animals (32, 33). There is suggestive epidemiolgica! evidence that nitrosa- either hospital based or have relied exclusively on surrogate interviews, the present study compares representative cases with general population controls and about one-half of the cases studied were interviewed directly. The magnitude of risk esti mates from use of tobacco and alcohol in this study are in general agreement with those of earlier studies. We found ex- smokers to have a reduced risk for carcinoma of the esophagus mines are involved in the etiology of esophageal carcinoma. Preformed nitrosamines and their precursors (nitrite, nitrate, and secondary amines) were detected in various common foods and drinking water consumed in Linxian County, China, an extremely high-risk area for esophageal cancer (34-39), and residents of Linxian were found to have significantly higher urinary levels of nitrate and several W-nitrosamino acids than 3846 TOBACCO, ALCOHOL, DIET, OCCUPATION, AND ESOPHAGEAL CARCINOMA residents of Fanxian, a nearby county with relatively low rates ACKNOWLEDGMENTS of esophageal cancer (40). Occupational exposure to metal dust, especially beryllium, was found to be a risk factor for carcinoma of the lower one- third of the esophagus. The esophagus is potentially exposed We thank John Mouzakis for conducting the interviews, Kazuko Arakawa for her assistance in data analysis, and Johannes Berkel, M.D., for his helpful suggestions. to inhaled dust particles since dust deposited in the respiratory tract is ultimately cleared by the mucociliary apparatus and is swallowed. A number of metals, including inorganic arsenic, chromium, and nickel, have been shown to be risk factors for respiratory cancers in humans (41, 42), lending some credibility to the hypothesis that the observed association between esoph ageal cancer and exposure to metal dusts is a causal one. Two studies have suggested that plumbers and pipefitters are at an increased risk of esophageal cancer (43, 44). We found a nonsignificant excess of plumbers and pipefitters among the cases. Six cases relative to one control were previously em ployed as plumbers or pipefitters (P = 0.13). All six cases were first employed as plumbers or pipefitters at least 25 years prior to the diagnosis of cancer. The duration of employment of these cases was 3 (2 cases), 5, 9, 22, and 35 years, respectively. In addition to metal dusts and fumes, plumbers and pipefitters are potentially exposed to known or suspected carcinogenic sub stances such as asbestos, tar, benzene, and chlorinated aliphatic solvents. There have been reports that populations heavily exposed to asbestos have experienced an increased risk of esophageal can cer (45, 46). We did not find an association between asbestos exposure and esophageal carcinoma risk. Although this study was population based, the participation rate was relatively low (56% of eligible patients). In addition, due to the rapidly fatal nature of the disease, only 47% of cases who participated in the study were interviewed directly. The generality and validity of our findings, therefore, needs to be critically evaluated. A comparison of the 275 cases studied with the 488 eligible cases showed the two groups to be comparable with respect to all items routinely collected by the Cancer Surveillance Program (sex, race, age, religion, place of birth, marital status, social class, subsite, and histology), and it seems reasonable to assume that the study cases are representative of eligible patients. A comparison of results based on all pairs with those based on only directly interviewed pairs found them to be consistent; the 95% confidence intervals associated with RRs from the two sets are largely overlapping. Interview status (direct versus proxy) does not appear to have affected the quality of the information collected. This is not entirely surprising since the information sought relates to past events and life-style factors familiar to spouses and first degree relatives. We also noted a high degree of consistency in the strength of the observed associations across sex-race strata, and this internal consistency of our data provides some assurance that these findings reflect real differences. Among black male controls, 53% were current smokers, 33% consumed more than 80 g of ethanol per day, 20% ate fried bacon or ham at least 5 times per week, 50% had fresh fruits and raw vegetables less frequently than 5 times a week, and 33% preferred whole grain bread. In contrast, the corresponding percentages among the white male controls were 37, 19, 8, 25, and 44%. Our study, therefore, suggests that differences in smoking, drinking, and dietary habits all contribute to the differential risk of esophageal carcinoma between United States blacks and whites. There is no evidence that occupational exposure to metal dust plays a role in this black-white risk difference. REFERENCES 1. Day, N. E., Muoz,N., and Ghadir-an, P. Epidemiology of esophageal cancer: a review. In: P. Correa and W. Haenszel (eds). Epidemiology of Cancer of the Digestive Tract, pp. 21-57. The Hague: Martinus Nijhoff Publishers, 1982. 2. Day, N. E. Some aspects of the epidemiology of esophageal cancer. Cancer Res., 35: 3304-3307, 1975. 3. Waterhouse, .1., Shanmugaratnam, K., Muir, C., and Powell, J. Cancer Incidence in Five Continents, Vol. 4. IARC Scientific Publication No. 42. Lyon, France: International Agency for Research on Cancer, 1982. 4. Blot, W. J., and Fraumeni, J. F. Trends in esophageal cancer mortality among U.S. blacks and whites. Am. J. Public Health, 77: 296-298, 1987. 5. Mack, T. M. Cancer Surveillance Program in Los Angeles County. Nati. Cancer Inst. Monogr., 47: 99-101, 1977. 6. Wright, W. E., Bernstein, L., Peters, J. M., Garabrant, D. H., and Mack, T. M. Adenocarcinoma of the stomach and exposure to occupational dust. Am. J. Kpidemini., in press, 1988. 7. Breslow, N. E., and Day, N. E. Statistical Methods in Cancer Research, Vol. 1. IARC Scientific Publication No. 32. Lyon, France: International Agency for Research on Cancer, 1980. 8. Bruzzi, P., Green, S. B., Byar, D. P., Brinton, L. A., and Schairer, C. Estimating the population attributable risk for multiple risk factors using case-control data. Am. J. Epidemiol., 122: 904-914, 1985. 9. Adams, D. F. Nutritive value of American foods in common units. Agriculture Handbook No. 456, United States Department of Agriculture, pp. 31. Wash ington, DC: United States Government Printing Office, 1975. 10. Rogot, E., and Murray, J. L. Smoking and causes of death among U.S. veterans: 16 years of observation. Public Health Rep., 95: 213-222, 1980. 11. Kahn, 11.A. The Dorn study of smoking and mortality among U. S. veterans: report on eight and one-half years of observations. Nati. Cancer Inst. Mon ogr., 19:1-125, 1966. 12. Hammond, E. C. Smoking in relation to the death rates of one-million men and women. Nati. Cancer Inst. Monogr., 19: 127-204, 1966. 13. Doll, K., and Peto, R. Mortality in relation to smoking: 20 years' observations on male British doctors. Br. Med. J., 2:1525-1536, 1976. 14. Weir, J. M., and Dunn, J. E. Smoking and mortality: a prospective study. Cancer (Phila.), 25.- 105-112, 1970. 15. Hirayama, T. Changing patterns of cancer in Japan with special reference to the decrease in stomach cancer mortality. In: H. H. Hiatt, J. D. Watson, and J. A. Winsten (eds), Origins of Human Cancer, pp. 55-75. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, 1977. 16. Hakulinen, T., Lehtimaki, L., Lehtonen, M., and Teppo, L. Cancer morbidity among two male cohorts with increased alcohol consumption in Finland. J. Nati. Cancer Inst., 52: 1711-1714,1974. 17. Monson. R. R., and Lyon, J. L. Proportional mortality among alcoholics. Cancer (Phila.), 36: 1077-1079,1975. 18. Schmidt, W., and Popham, R. E. The role of drinking and smoking in mortality from cancer and other causes in male alcoholics. Cancer (Phila.), 7:1031-1041, 1981. 19. Phillips, R. L. Cancer among Seventh-Day Adventists. J. Environ. Pathol. Toxicol., 3: 157-169, 1980. 20. Enstrom, J. E. Cancer mortality among Mormons in California during 196875. J. Nati. Cancer Inst., 65:1073-1082, 1980. 21. Hirayama, T. Diet and cancer. Nutr. Cancer, /: 67-81, 1979. 22. Wynder, E. L., and Bross, I. J. A study of etiological factors in cancer of the esophagus. Cancer (Phila.), 14: 389-413, 1961. 23. Pottem, L. M., Morris, L. E., Blot, W. J., Ziegler, R. G., and Fraumeni, J. F. Esophageal cancer among black men in Washington, D. C. I. Alcohol, tobacco, and other risk factors. J. Nati. Cancer Inst., 67: 777-783, 1981. 24. Martinez, I. Factors associated with cancer of the esophagus, mouth, and pharynx in Puerto Rico. J. Nati. Cancer Inst., 42: 1069-1094, 1969. 25. Vassallo, A., Correa, P., De Stefani, E., Cendan, M., Zavala, D., Chen, V., Carzoglio, J., and Deneo-Pellegrini, H. Esophageal cancer in Uruguay: a case-control study. J. Nati. Cancer Inst., 75: 1005-1009, 1985. 26. Victora, C. G., Muoz,N., Day, N. E., Bartolos. L. B., Peccin, D. A., and Braga, N. M. Hot beverages and esophageal cancer in southern Brazil: a case-control study. Int. J. Cancer, 39:710-716, 1987. 27. Tuyns, A. J., Pequignot, G., and Abbatucci, J. S. Le cancer de l'oesophage en Ilk- ci Vilaine en fonction des niveaux de consommation d'alcool et de tabac. Des risques qui se multiplient. Bull. Cancer (Paris), 64:45-60, 1977. 28. De Jong, U. W., Breslow, N., Goh Ewe Hong, J., Sridharan, M., and Shanmugaratnam, K. Aetiological factors in oesophageal cancer in Singapore Chinese. Int. J. Cancer, 13:291-303, 1974. 29. Tuyns, A. J., Riboli, E., Doornbos, G., and Pequignot, G. Diet and esophageal cancer in Calvados (France). Nutr. Cancer, 9: 81-92, 1987. 30. DeCarli, A., Liati, P., Negri, E., Franceschi, S., and La Vecchia, C. Vitamin A and other dietary factors in the etiology of esophageal cancer. Nutr. Cancer, 10:29-37, 1987. 31. Stephany, R. W., Freudenthal, J., and Schuller, P. L. Quantitative and 3847 TOBACCO, ALCOHOL, DIET, OCCUPATION, AND ESOPHAGEAL CARCINOMA qualitative determination of some volatile nitrosamines in various meat products. IARC Sci. Pubi., 14: 343-360, 1976. 32. Survey of Compounds Which Have Been Tested for Carcinogenic Activity. NIH Publication No. 80-453, Washington, DC: U. S. Department of Health, Education, and Welfare, 1978. 33. Stinson, S. F. Esophageal carcinoma, Model No. 190. In: C. C. Capen, D. B. Hackel, T. C. Jones, and G. Migaki (eds.). Handbook: Animal Models of Human Disease, Fascicle 9. Washington, DC: Registry of Comparative Pathology, Armed Forces Institute of Pathology, 1980. 34. Co-ordinating Group for Research on the Etiology of Esophageal Cancer of North China. The epidemiology of esophageal cancer in north China and preliminary results in the investigation of its etiological factors. Sci. Sin., 18: 131-148, 1975. 35. Li, M. H., Lu, S. H., Ji, C, Wang, M. Y., Cheng, S. J., and Jin, C. L. Formation of carcinogenic A'-nitroso compound in corn-bread inoculated with fungi. Sci. Sin., 22: 471-477, 1979. 36. Lneuw, SjV. -Hni.t,roLsio, cMom. pHo.,unJdi,, CA.',-3W-maentgh,ybMut.ylY-A.,r-Wmaenthgy, laYce. toLn.,yla-nnditroHsuaamnign,e, L. iAn corn-bread inoculated with fungi. Sci. Sin., 22: 601-608, 1979. 37. Wang, G. H., Zhang, W. X., and Chai, W. G. The identification of natural Roussin red methyl ester. Acta Chem. Sin. 38:95-102, 1980. 38. Croisy, A., Ohshima, H., and Bartsch, H. Nitrosating properties of bismethyl-thio-diiron-tetranitrosyl (Roussin's red methyl ester), a nitroso com pound isolated from pickled vegetables consumed in northern China. IARC Sci. Pubi., 57: 327-335, 1984. 39 Wang, Y. L., Lu, S. H., and Li, M. M. Dterminationof nitrates and nitrites in well water from Yaocun Commune: Lin-xian County, Henan Province. Chin. J. Oncol., 1: 201-205, 1979. 40 Lu, S-H., Ohshima, H., Fu, H-M., Tian, Y., Li, F-M., Blettner, M., Wahren- dorf, J., and Bartsch, H. Urinary excretion of .Y-niirosaminn acids and nitrate by inhabitants of high- and low-risk areas for esophageal cancer in northern China: endogenous formation of nitrosoproline and its inhibition by vitamin C. Cancer Res., 46:1485-1491, 1986. 41 International Agency for Research on Cancer Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans. Some Metals and Metallic Compounds, Vol. 23. Lyon, France: International Agency for Research on Cancer, 1980. 42. Fraumeni, J. F. Respiratory carcinogenesis: an epidemiologie appraisal. J. Nati. Cancer Inst., 55: 1039-1046, 1975. 43. Kaminski, R., Geissert, K. S., and Dacey, E. Mortality analysis of plumbers and pipefitters. J. Occup. Med., 22: 183-189, 1980. 44. Milham, S. Occupational Mortality in Washington State, 1950-1979. DHHS (NIOSH) Publication No. 83-116. Washington, DC: United States Depart ment of Health and Human Services, 1983. 45. Selikoff, I. J., Hammond, E. C., and Seidman, H. Mortality experience of insulation workers in the United States and Canada, 1943-1976. Ann. NY Acad. Sci., 330: 91-116, 1979. 46. McDonald, J. C., Liddee, F. D. K., Gibbs, G. W., Eyssen, G. E., and McDonald, A. D. Dust exposure and mortality in chrysotile mining, 19101975. Br. J. Indust. Med. 37:11-24, 1980. 3848