Tobacco Documents Online

Environmental Factors and High Risk of ophageal Cancer Among Men in Coastal South Carolina Linda Morris Brown,* William J. Blot, .Stanley H. Schuman, Vard M. smith, Abby G,"Ershow Richard..D Marksl " Joseph F. Fraumeni, Jr. ".~ ease-cOntrol study involving, interviews of" 207 men.with esophageal cancer and 422 control subjects or their next of kin was conductedto identify reasons for the unusually high rates of esophageal cancer among men in coastal South Carolina. Tobacco and alcohol, including moonshine, were identified a~. the major determinants of esophageal cancer risk. Increased risk was also associated with low intake of fresh fruits but not with drinking of local herbal teas. The findings suggest that efforts aimed at reducing tobacco and alcohol use will help to lower the elevated rates of esophageal cancer in coastal South Carolina. [J Natl Cancer Inst 1988;80:1620-1625] A survey of mortality in the 1940s (1) identified twofold to threefold elevations in the rates of esophageal cancer • among black men living in the South Carolina low country (an area of islands and coastal mainland in the vicinity of Charleston), compared with rates in the state as a whole. Subsequent mortality data have shown that excessive rates have persisted; some of the highest rates of esophageal cancer in the nation (two to three times the national average) are • found among blacks in counties along the South Carolina coast (2,3). We report results from a case-control study evaluating the possible determinants of esophageal cancer in black and white men in this high-risk area. Subjects and Methods Subjects The case-control study consisted of two components: a hospital-based incidence study (incidence series) enrolling patients during 1982-1984 and a next-of-kin mortality study (mortality series) covering deaths during 1977-1981. Con- trol subjects were excluded from consideration if the .diag- nosis at admission or the cause of de~th v~as related to alco- hol and/or diet, two of the major factors under investigation. I Thus, diagnoses of cancer (except lymphoma and leukemia), benign oral. tumors, 0ral and esophageal diseases, liver' cir- rhosis, and nutritional deficiencies resulted in exclusion. In addition, death by suicide or homicide and diagnosis of a mental disorder resulted in exclusion. Incidence Series ' ~ Cases of primary esophageal cancer for the incidence se- ries were identified through radiotherapy departments and pathology records at four Charleston hospitals by R. D. Marks, a radiotherapist who helped to obtain data from pa- tients at these collaborating centers (Medical University of South Carolina, Charleston Memorial Hospital, Roper Hos- pital, and the Veterans Administration Hospital)• These re- ferral centers treat most of the patients with esophageal can- cer in the coastal area. Additional cases were found through area surgeons whose patients were receiving treatment other than radiotherapy. For entry in the study, men were required to be South Carolina residents --<79 years old at the time of hospital admission or treatment (January 1, 1982, through October 31, 1984). Two control patients per case were iden- tified through admission records at the same hospital in the same time period. They were also similar to the case patients with respect to race and age (:t:5 yr). Mortality Series Subjects for the mortality series were male residents of eight coastal counties of South Carolina (Beaufort, Berkeley, Charleston, Colleton, Dorchester, Georgetown, Horry, and Jasper) who had died of primary esophageal cancer during 1977-1981 at the age of-<79 years. Two control subjects for Recei,~d May'2~, 1988; revised Septembe~ 21•, 1~88; accepted September 29, 1988. L. M. Brown, W. J. Blot, A. G. Ershow, and J. F. Fraumeni, Jr., Epir demiology and Biostatistics Program, Division of Cancer Etiology, National Cancer Institute, Bethesda, MD. S. H. Schuman, V. M. Smith, and R. D. Marks, Medical University of South Carolina, Charleston, SC. We thank Jack Cahill, Marcy Breslow, Diane Tate, and Susan West for administrative assistance, Nancy Bayless and Doug Midthune for computer. support, Gloria Gridley and Dr. Regina Ziegler for technical assistance, Cherry Seabrook for interviews of study subjects, Amilda Bayne for sec- retarial support, and Murray Hudson for death certificate infor.ma3ion. *Correspondence to:. Linda Morris Brown, Executive Plaza North, Rm. 415, National Institutes of Health, Bethesda, MD 20892. 1620 Journal of the National Cancer Institute

dach cage were randomly selected to be similar with respect tq race, age, county of residence, and year of death. Cases and controls were identified by the Office of Vital Records and Public Health Statistics, South Carolina Department of ~Haealth and Environmental Control. Copies of death certifi- tes were obtained to verify the cause of death and to aid in tracing the next of kin. Questionnaire All patients in the incidence series were interviewed di- rectly in the hospital, whereas in the mortality' series, the next of k~n ~(usually a. spo.use..or.c.los.e r.e.lativ.e) were inte.rview.ed • -at home. A structured questi0nnairewa~ u~ed to •obtain in- formation on usual adult use of tobacco, alcohol, coffee, and tea (including herbal teas); usual adult diet; medical and den- tal history; usual occupation; familial occurrence of cancer; and several demogra.phic x~ariables. .Assessmen.t of Risk iFactors. Alcohol intake. A drinker was defined as one who had at least one drink of bee~:, wine, hard liquor, or moonshine per week for _>1 year. Usual adult consumption of beer, wine, hard liquor, and moonshine was ascertained separately for • the weekend and for the rest of the week. For drinkers, weekly consumption of each type of alcoholic beverage was calculated as the sum over these two periods. We estimated average weekly ethanol intake assuming that 1 fluid ounce of beer, wine, or hard liquor yields !.1, 2.9, or 9.4 g of ethanol, respectively (4). Total ethanol consumption was calculated by summing the ethanol contribution from beer, ine, and hard liquor but not from moonshine, which has ~known and possibly varying ethanol .content. For ease of interpretation, this sum was then converted into hard liquor equivalents (fluid ounces per day). Three levels of consumption (light, moderate, and heavy) were designated by dividing the frequency distribution of consumption for the control subjects into approximate thirds. Tobacco use. A tobacco user was def'med as one who smoked >_100 cigarettes during a lifetime or used at least one cigar, pipe, pouch or plug of tobacco, or small can of snuff p.er week for ->1 year. Estimates of usual tobacco use were based on the reported usual number of cigarettes smoked per day, pipes and cigars smoked per week, pouches of tobacco chewed per week, or cans of snuff dipped per week during most of a lifetime and the number of years each was used. We also ascertained the number of years each type of tobacco was used, the age at which cigarette smoking was started and stopped, and whether filtered or nonfiltered cigarettes were usually smoked. Diet. The dietary section of the questionnaire sought in- formation on usual adult, consumption of 65 food items. Frequencies of consumption of food from food groups such as citrus fruits and juices (oranges, grapefruit, orange juice, grapefruit juice, and lemonade) were calculated by summing the frequencies for the appropriate food items. Specific nu- ~ulervnt and calorie intakes were estimated with an algorithm ed on recent National Health and Nutrition Examination ey data (5,6). Three consumption categories (low, mod- erate, and high) were created for each food group and nu- trient index by dividing the frequency distribution for the control subjects into approximate thirds. If information was provided by a respondent other than the subject or his wife, the subject was excluded from the dietary analyses because of the potential for lack of sufficient familiarity with the in- dividual's diet and greater likelihood of misclassification of food and nutrient intake. Statistical Analysis Unconditional binary logistic •models were used to simul- taneously estimate the effects of the smoking, drinking, and other exposure va.riables ..(7,~). Categories. of Cigarett~ use. "(-<!9, ~20-29, o~: %_30 cigarettes per da3~), ethan'~)l c0nsumP= tion (_<3.0, 3.1-9.0, or >9.0 oz/day), moonshine use (yes or no), and study series (incidence or mortality) were included in models to adjust for potential confounding. Addition of the matching factors (age, race, hospital, and county of res- idence) to the logistic models resulted in little ch.ange in the.. • estimate~. T6 tbst for linear trend, a categorical var]~ble was entered as a continuous variable in the logistic model. Sum- mary odds ratio (OR) estimates for the dietary variables ad- justed for use of cigarettes (-<19 or -->20 cigarettes per day) and ethanol (_<9 or >9.0 oz/day) were obtained by max- imum likelihood procedures (9,10). Mantel's extension test (11) was used in a two-tailed test for trends in risk associated with frequency of intake of the dietary variables. Results The potential study sample for the incidence series con- sisted of 96 hospitalized patients with esophageal cancer and 173 control patients. There were 87 hospitalized patients with esophageal cancer who met the eligibility requirements, and interviews were completed for 74 (85%). There were 165 eligible control patients, and interviews were completed for 157 (95%). The reasons for not interviewing the patients were physician refusal (six cases and one control); patient re- fusal (two cases and six controls); death (two cases); patient incompetence (one control); and hospital discharge prior to interview (three cases). For the mortality series, 143 subjects who died during 1977-1981 from primary esophageal cancer met the eligi- bility criteria. There were 285 eligible control patients. In- terviews with the next of kin were completed for 401 (94%) of the subjects in the mortality series. The primary reasons for not interviewing the next of kin were refusal of the re- spondent (5%) and inability to locate the (espondent (1%). The final study population consisted of 207 case subjeet~ (159 blacks and 48 whites) and 422 control subjects (324 blacks and 98 whites), after excluding three subjects of un- known or other race. For cases in the incidence series, the histologic types of esophageal cancer were squamous cell carcinoma (85%), adenocarcinoma (13%), and carcinosar- coma (2%). The distribution of cancer sites in the esophagus was as follows: proximal (4%); proximal and middle (13%); middle (36%); middle and distal (3~); distal (29%); gastro- esophageal junction (13%); entire esophagus (1%); and not ascertained (1%). Similar detailed information on histology and site was not ascertained for the cases in the mortality se- ries; therefore, the number of histologically confirmed cases Vol. 80, No. 20, December 21, 1988 ARTICLES 1621

of cell types other than squamous cell carcinoma was too small to allow an investigation of risk factors by histologic type. The admission diagnoses for the hospital control pa- tients included circulatory disease (22%); respiratory disease (13%); genitourinary disorders (12%); and several other dis- eases, each of which accounted for no more than 10%. The causes of death for control subjects in the mortality series were heart disease (42%), cerebrovascular and other circula- tory diseases (21%), respiratory disease• (9%), accidents (8%5, artd several~.pther less common Causes. The me.dian age (62 • yr) bf ihe s.ubjerts w..~s .the same f6r ~a~es and contr~lg'~ and >85% of the subjects had spent most of their childhood' and adulthood in South Carolina. The mean number of years of schooling completed by the subjects was 6.8 for the cases and 7.2 for the controls. Tobacco ...... In the combined study population, only three (1.4%) of the case subjects h~/d not used some form of tobac.co, and 59 (14%) of the.control subjects were nonusers (table 1); hence, the ORs associated with each form of tobacco use were high despite a wide 95% confidence interval (CI). Cigarette smoking was the most common form of tobacco use, but there were large significant increases in risk (OR >_ 7.7) for all forms of use except exclusive smokeless tobacco use. Because the pattern of cigarette use and the magnitude of smoking-related risks were similar for the incidence series and the mortality series, risks for the combined study popu- lation are combined in table 2. Risks increased significantly with both intensity (P for trend = < ,0001) and duration (P for trend = .03) of cigarette use. Risks for men who had stopped smoking cigarettes for _>10 years were similar to risks for those who had never smoked. There was not a great difference in risk for men who usually smoked non- filter cigarettes, compared with risk for those who usually smoked filter cigarettes. The adjusted ORs for nonfilter ver- sus filter smoking were 1.6 (95% CI, 0.7-3.7) and 0.9 (95% CI, 0.5-1.6) for the incidence series and the mortality series, respectively. Alcohol In the combined study population, only 15 (7%) of the case subjects had not consumed alcoholic beverages, corn- Table'l, Risks .of esophageal cancer according to tpbacco use No. of -- Crude Adjusted* Tobacco use OR Cases Controls OR 95% CI Nonuser 3 59 1.0 1.0 -- Smokeless tobacco 1 12 1.7 1.2 0.1-13.3 only Pipe and/or cigar only 14 22 13.1 9.9 2.5-38.5 Pipe/cigar/smokeless 7 8 16.3 17.1 3.6-81.4 tobacco only Cigarettes only 134 242 10.3 7.7 2.3-25.5 Other combinations 48 79 10.9 8.5 2.5-29.2 * All risks are relative to risk for nonusers of tobacco and are adjusted for study series (incidence or mortality) and use of ethanol in a logistic model. Table 2, Risks of esophageal cancer according to cigarette smoking characteristics No. of- Adjusted~" Characteristic Crude Cases* Controls* OR OR 95% CI No. of cigarettes/day 0 25 1-19 28 20-29 89 -->30 63 No. of yr of smoking • 0 25 1-24'.." .'".':.. 17. .25-44 100 -->45 59 Smoking status Nonsmoker 25 Current smoker 128 . Stopped 1-9 yr ago 34 Stopped ->10 yr agp 8 101 1.0 1.0 -- 111 1.0 0.8 0.4-1.5 129 2.8 2.0 1.1-3.4 74 3.4 2.6 1.4-4.7 101 1.0 1.0 ' ' .~:3 " .1.6" "" 1;4. 177 2.3 1;6 '1.0-2.8 93 2.6 " 1.8 1.0-3.3. 101 . 1.0 1.0 -- 202 2.6 1.8 1.0-3.0 55 2.5 2.0 1.0-3.7 63 1.2 1.0 0.5-2.1 *Cases an~ cbntrols with unknown value~ for ~mot~nt (two cases and" seven controls), duration (six cases and eight controls), or cessation (two cases and one control) of smoking are excluded. ~fAll risks are relative to risk for nonsmokers for each smoking characteristic and are adjusted for study series (incidence or mortality) and use of moonshine and other ethanol in a logistic model. pared with 95 (23%) of the control subjects. Most of the men reported consumption of several types of alcoholic bev- erages over a lifetime. Increased risks of esophageal cancer were associated with all types of alcoholic beverages, al- though the OR for consumption of beer and/or wine only was only slightly elevated (table 3). Adjustment for cigarette use reduced the magnitude of the associations with alcohol to some extent, but the ORs ranged from 2 to 4 for drinkers of liquor only, moonshine only, and both liquor and moonshine. There were significant trends of increasing risk of esophageal cancer with increasing intake of ethanol from beer, wine, and hard liquor consumed for subjects in both the incidence series (P = .002) and the mortality series (P = .01) (table 4). Although crude risks were substantially higher for the incidence series, adjustment for use of cigarettes and moonshine reduced these risks and brought them closer to those for the mortality series. Because the number of nonusers was too small, it was not possible to evaluate the risk of drinking in the absence of smoking and vice versa or Table 3. Risks of esophageal cancer according to alcoholic beverage use ~ No. of --~ Crude Adjustedt Alcoholic beverage use . OR ~ ~ ~Cases* Controls* OR 95% C~[ Nondrinker 15 95 1.0 1.0 -- Beer and/or wine only 11 50 1.4 1.1 0.4-2.6 Hard liquor only 20 41 3.1 2.1 1.0-4,8 Moonshine only 5 9 3.5 2.6 0.7-9.6 Moonshine and hard 25 32 4.9 3.7 1.7-8.2 liquor only Other combinations 129 190 4.3 3.0 1.6-5.7 *Two cases' and five controls for which drinking habits were unknown are excluded. ~ All risks are relative to risk for nondrinkers of alc.rholic beverages and are adjusted for study series (incidence or mortality) and use of cigarettes in a logistic model. 1622 Journal of the National Cancer Institute

Table 4. Risks of esophageal cancer by studyseries according to ethanol consumption Series Ethanol* No. of- Crude Adjusted:l: (oz/day) Cases]" Controls]" OR OR 95% CI Incidence 0 3 29 1.0 1.0 -- 0. 1-3.0 " 6 44 1.3 0.9 0.2-4.3 3.1-9.0 16 38 4.1 1.7 0.4-7.3 >9.0 47 45 10.1 3.6 0.9 - 14.9 Mortality 0 17 75 1.0 1.0 -- 0.1-3.0 31 63 .22 1.9 0.9r3.8. *Ethanol from moonshine was not included in this ethanol estimate. ]'Two hospital cases, 19 mortality series cases, and 30 mortality series controls for which ethano! consumption was unknown are excluded. ~:AII risks are relative to risks for nondrinkers of ethanol for each study series and are adjusted for use of cigarettes and moonshine in a logistic model. to adequately assess the interactive effects of smoking and drinking. In the incidence series, there were significant trends of increasing risk of esophageal cancer with increasing con- sumption of moonshine (P = .004) (table 5). A similar trend was not seen for the mortality series (P = .32). Of the case patients in the incidence series, 78% reported regu- lar use of moonshine, with blacks reporting higher use than whites (85% of blacks and 53% of.whites). Of the control patients in this series, 49% of the blacks and 19% of the whites reported regular consumption of moonshine. Moon- shine drinkers tended to be drinkers of hard liquor, and heavy drinkers of moonshine were also likely to be heavy drinkers of hard liquor, but risks associated with moonshine consump- tion in the incidence series remained significantly elevated even after adjustment for smoking and ethanol intake from beer, wine, and hard liquor. Herbal Teas Consumption of several local teas as home remedies was commonly reported. Sassafras tea was used by more than two-thirds of the control subjects, and "pinetop" and "rabbit tobacco" (molasses) teas were used by ~40%. Use of sweet gum, boneset bush, and "dark root" teas or other herbal teas Table 5. Risks of esophageal cancer by study series according to moonshine consumption Series Moonshine No..of -- Crude Adjusted]' .... (oz/day) Cases* • Controls* OR OR 95% CI Incidence Mortality 0 16 89 1.0 1.0 -- 0.1-2.1 6 20 1.7 1.9 0.6-5.8 2.2-6.9 18 22 4.6 3. I 1.2-7.9 ___7.0 34 25 7.6 3.4 1.4-8.0 0 76 185 1.0 1.0 -- 0.1-2.1 18 27 1.6 1.6 0.8-3.2 2.2-6.9 15 18 2.0 1.7 0.8-3.7 _>7.0 12 21 1.4 1.0 0.4-2.3 Twelve mortality series cases and 15 mortality series controls oonshine consumption was unknown were excluded. for which ]" All risks are relative to risks for nondrinkers of moonshine for each study series and are adjusted for use of cigarettes and ethanol in a logistic model. was reported by 3%-8% of the control subjects. Consumption by case subjects was generally comparable; the tobacco- and alcohol-adjusted ORs for consumption of sassafras, pinetop, and rabbit tobacco teas were 0.8 (95% CI, 0.4-1.5), 1.5 (95% CI, 0.8-2.8)~ and 1.0 (95% CI, 0.5-1.8), respectively. The ORs for the less commonly used herbal teas were <_1.0. No clear trends were associated with drinking regular tea (either hot or iced) or coffee. Risks were not elevated for those who drank tea without milk or cream or for those who drank coffee or tea. "very hot." Diet ...... :"'::"" "' "" : "~" After adjustment for smoking a~d "drinl~ing, significantly increased risks of esophageal cancer were associated with low intake of fruits, particularly citrus fruits and juices, and with high intake of liver (table 6). Approximately twofbld increases in risk were seer~ for.. subject.s wit.h the highest. compared with the lbwest intake of retinol (table 7). ORs for vitamin C and fiber for subjects with the highest intake were approximately one-half of ORs for subjects with the lowest intake. There was no risk associated with usual adult relative body weight (weight/height2), although risks were slightly higher for subjects with lower intake of calories (including food but not alcohol) and for those who usually ate two or fewer meals per day, compared with those who ate three or more meals per day. There were no marked case- control differences associated with the method of preparing or cooking food. Controlling the analyses of smoking and drinking for intake of fruits or nutrient indices yielded litde change in the ORs for those variables. Other Factors Risk of esophageal cancer was not associated with em- ployment in administrative/sales, service, farm, mechanics, Table 6. Association of esophageal cancer risk with frequency of intake of food from food groups OR for frequency of intake (95% el)* Food group Low Moderate High Dairy products 1.0 1.0 (0.6-1.8) 0.6 (0.4-1.2) Eggs1.0 0.6 (0.3-1.2) 0.7 (0.4-1.2) Meat and poultry 1.0 1.5 (0.9-2.7) 1.1 (0.6-2.0) Fish and shellfish 1.0 1.2 (0.7-2.2) 1.2 (0.6-2.1) Liver 1.0]. 2.4 (1.3-4.7) 2.2 (1.1-4.3) All vegetables 1.0 • 1.1 (0.6-1.8) 0.7 (0.4-1.3) All fruits 1.0:~ 0.5 (0.3-1~.9) 0.5 (0.3-0.9) Citrus fruits and juices 1.0:1: 0.5 (0.3-0.9) 0.5 (0.3-0.9) High-beta-carotene fruits 1.0 0.7 (0.4-1.3) 0.8 (0.5-1.5) Other fruits 1.0 0.9 (0.5-1.5) 0.6 (0.3-1.0) Tomatoes 1.0 1.1 (0.6-2.0) 0.7 (0.4-1.4) Potatoes L0 0.9 (0.5-1.5) 0.7 (0.4-1.4) Grains 1.0 0.5 (0.3-1.0) 1.0 (0.6,1.7) * All risks are relative to risk for subjects with low intake of each food group and are adjusted for use of cigarettes and alcohol in a stratified analysis. Risks are for combined study population; information was obtained from subject or his wife. Low --- lower third, moderate = middle third, and high = upper third of study population according to frequency of intake. ]"P for trend = <-05. ~:P for trend = %01. 0 O~ Vol. 80, No. 20, December 21, 1988 ARTICLES 1623

Table 7. Association of esophageal cancer risk with freq.uency of intake of nutrients ,lutrient OR for frequency of intake (95% CD* Low Moderate High Carotene 1.0 0.9 (0.5-1.5) 0.8 (0.5-1.6) Retinol 1.0~ 2.1 (1.1-3.9) 1.9 (1.0-3.5) Vitamin C 1.0~" 0.8 (0.4-1.3) 0.5 (0.3-1.0) Riboflavin 1.0 1.4 (0.8-2.5) 1.0 (0.5-1.8) Thiamine 1.0 1.0 (0.6-1.7) 0.6 (0.3-1.1) Folate 1.0 1.1 (0.6-1.9) 0.7 (0.4-1.3) Fiber 1.0~" 0.7 (0A-.I.2) 0.5 (0.3-1.0) *All risks are relati:ce to risk for subjects with low intake of each nutrient and are adjusted for use of cigarettes and alcohol in a stratified analysis. Risks are for combined study population; information was obtained from subject or his wife. Low = lower third, moderate = middle third, and high = upper third of study population according to frequency of intake. ~'P for trend = <.05 construction, production, 'transportation, helper, or military occupations. There was 6nly a weak, nonsignificant associ- ation with the number of years of schooling completed. His- tory of esophageal cancer in a first-degree relative, number of teeth lost, and use of mouthwash were not associated with an excess risk of esophageal cancer. Discussion This epidemiologic investigation hasidentified tobacco and alcohol as the primary determinants of esophageal can- among men in coastal South Carolina, with some con- from diets low in fresh fruits. Our .findings are thus similar to those reported from other areas of the United States, where smoking and drinking have been shown to ac- count for most esophageal cancers and nutritional factors have been implicated (12-15). One distinguishing feature of coastal South Carolina, how- ever, is the high prevalence of moonshine use, particularly among blacks. National data on moonshine drinking are not available for comparison, but the use of moonshine in South Carolina is probably considerably greater than the norm for th~ United States. Most of the subjects who were interviewed directly (incidence series) reported using moonshine, but the next of kin in the mortality series reported considerably less use of moonshine by the subjects because they lacked knowl- edge of their relatives' drinking habits. The resulting misclas- sification was probably'the reason that the.strong a.ssociation between moonshine use and esophageal cancer was seen only in the incidence series. Risks for patients in the incidence series rose with increasing consumption of moonshine, and the trend persisted even after we controlled the analysis for smoking and consumption of other alcoholic beverages. Al- though most moonshine drinkers consumed other forms of alcohol, risks were increased among those who drank only moonshine. The situation in South Carolina thus resembles that in high-risk areas of Puerto Rico (I6) and northern nce (17), where an association between esophageal can- and consumption of home-brewed alcoholic beverages (rum and apple cider whiskies, respectively) has been re- ported. As in several other studies of esophageal cancer (18-24), we found evidence of dietary effects; increased risks were associated with low intake of fruits during adulthood. The effects of low fruit consumption persisted after we controlled the analysis for use of alcohol and tobacco. The strongest association was with citrus fruits and correspondingly with vitamin C intake. Previous studies (21,22) identified a link between esophageal cancer and poor diet, but our finding that higher intake of fruits, especially citrus fruits, reduces the risk of esophageal cancer is consistent with. the results of recent studies in France (23). and...Italy (24)~ .It is not de~:;, howefer, whether th~ reduction i.n .esophageal cancer. risk with higher intake of.fruits and citrus fruits' is du~ to' ascorbic acid or to some other constituent ot citrus fruits. A diet low in ascorbic acid, which blocks the endogenous formation of N-nitroso compounds (25), is postulated to be involved in the etiology of esophage~/l cancer in ~ome .high-risk. are.~s o£ the word (26):. Smoking, ~a known' risk factor for esophageal cancer, may compound the problem, since the amount of vitamin C needed to achieve steady-state plasma concentrations is ~40% greater in smokers than in nonsmokers (27). Unfortunate/y, the relatively small number of nonsmokers in our study precluded investigation of this effect. We detected a positive association between retinol'intake and esophageal cancer risk. Natural and synthetic retinoids have been shown to inhibit chemically induced tumors in ex- perimental animals, but some investigators using carcinogen bioassays, including one in nitrosamine-induced esophageal cancer in rats, have reported tumor enhancement following retinol administration (28-30). We are not proposing that retinol intake increases esophageal cancer risk, but it is of interest that positive associations were also seen in studies in France (23) and Italy (24). In addition, higher consump- tion of liver, the food that contributed >76% of the retinol consumed by subjects in our study, was found in studies of esophageal cancer conducted in Washington, DC (22), France (23), and Italy (24). The liver has a high capacity to bind, metabolize, and excrete chemicals. In addition, the liver and the kidney probably concentrate more toxicants than any other organs (31). Therefore, it is conceivable that consumption of liver and chemicals stored in the liver may be carcinogenic to the esophagus. The interviews included questions about consumption of several herbal teas indigenous to the coastal area. It has been hypothesized that there is an association between el- evated rates of esophageal cancer in residents of South Car- olina and drinking of teas from local plants, which may contain high levels of tannin, safrole, or other agents. Re- suits of animal experiments have suggested that these sub- stances may act as carcinogens or cancer-promoting agents (32-34). Other studies have suggested that the high tannin content of sorghum and of mate has contributed to elevated rates of esophageal cancer in parts of Africa (35,36) and South America (37), respectively. In addition, drinking of hot tea gruel has been implicated as a possible risk factor for esophageal cancer in Japan (38,39) and in high-risk ar- eas of the Soviet Union and India (18,40). While we did find consumption of several herbal teas in South Carolina Journal of the National Cancer Institute

to be common, intake was generally not.~more frequent in case subjects than in control subjects. Furthermore, those who drank regular tea "black" were not at higher risk than those who drank it with milk or cream, which bind tannin and reduce its absorption. Thus, it is unlikely that ingestion of teas or local plant products exerts any substantial influence on esophageal cancer risk. Drinking tea at exceptionally hot temperatures has long been considered to be a potential risk factor in other populations (!8,19), and thermal injury from mate drinking is thought to con~tribute to both precancerous lesions and esoph.ageal cancer, in high:risk areas of South America (20,41), but we found no risk associated frith con- sumption of hot beverages. The results of this investigation suggest that use of cig- arettes and alcohol are the major risk factors for esophageal cancer among men in' th6 South Carolina low country and that diets low in fresh fruits, especially citrus fruits, also enhance risk. Although further research is necessary to eval- uate the carcinogenic!ty of moonshine, it seems clear that efforts to reduce tobacco and alcohol use and perhaps to encourage increased fruit consumption will help to lower the elevated rates of esophageal cancer in coastal South Carolina. References 1. BATES DC, CAs'ro~ JC, OBRLEN P, et al. Carcinoma of the esophagus in South Carolina. J SC Med Assoc 1971;11:453-456. 2. FRAUMEI, n JF, BLOt WJ. Geographic variation in esophageal cancer mortality in the United States. J Chronic Dis 1977;30:759-767. 3. RIGGAN WB, VAN BRUGGEN J, ACQUAVELLA JF, et al. US Cancer Mortal- ity Rates and Trends; 1950-1979. Vol I. NCI/EPA Interagency Agree- ment on Environmental Carcinogenesis. Washington, DC: US Govt Print Off, 1983 [EPA publication No. 60011-83-015a]. • 4. US Dept. of Agriculture. Foods commonly eaten by individuals: amount per day and per eating occasion. In: Home Economics Research Report No. 44. Washington, DC: US Govt Print Off, 1972. 5. BLOCK G, DRESSER CM, HARTMAN AM, et al. Nutrient sources in the American diet: quantitative data from the NHANES II Survey. I. Vita- mins and minerals. Am J Epidemiol 1985;122:13-26. 6. BLOCK G, DRESSER CM, HARTMAN AM, et al. Nutrient sources in the American diet: quantitative data from the NHANES II Survey. II. Macronutrients and fats. Am J Epidemiol 1985;122:27-40. 7. Cox DR. The analysis of binary data. London: Methuen, 1970:14-19. 8. DIXON WJ, ed. BMDP statistical software 1985. Berkeley, CA: Univ California Press, 1985:330-344. 9. NIART1N EJ, PICKLE LW. Odds ratio estimation for polychotomous ex- posures. Presented at the Fifth Annual Conference of the SAS User's Group International, Cary, NC, 1980. 10. GART JJ. The comparison of proportions: review of significance tests, confidence intervals, and adjustments for stratification. Rev Int Stat Inst 1971;39:148-161. 11. MANTEL N: Chi-square tests with one degree of freedom, extensions of Mantel-Haenszel procedure. J Am Stat Assoc 1963;58:690-700. 12. WYNDER EL, BROSS IJ. A study of etiological factors in cancer of the esophagus. Cancer 1961;14:389-413. 13. WYNDER EL, STELLMAN SD. Comparative epidemiology of tobacco- related cancers. Cancer Res 1977;37:4608-4622. 14. METTL~N C, GRAHAM S, PPaORE S, et al. Diet and cancer of the esoph- agus. Nutr Cancer 1981;2:143-147. 15. POTTERN LM, MORRIS LE, BLOT WJ, et al. Esophageal cancer among black men in Washington DC. I. Alcohol, tobacco, and other risk factors. JNCI 198 l;67:777-783. 16, MARTiNEZ I. Factors associated with cancer of the esophagus, mouth, and pharynx in Puerto Rico. JNCI 1969;42:1069-1094. 17. TUYNS AJ, PEQUtNOT G, ABBATUCCI JS. Oesophageal cancer and al- cohol consumption: importance of type of beverage. Int J Cancer 1979;23:443-447. t8. DAY NE, MUNOZ N. Esophagus. In: Schottenfeld D, Fraumeni JF Jr, eds. Cancer epidemiology and prevention. Philadelphia: Saunders, 1982:596-632. : 19. DEJONG UW, BR.ESLOW 'N, HONG G, e~t al. Aetiological factors in esophageal cancer in Singapore Chinese'. Int J Cancer 1974;13:~91- 303. 20. VICTORtA CG, MUNOZ N, DAY NE, et al. Hot beverages and esophageal cancer in southern Brazil: a case-control study. Int J Cancer 1987;39:710-716. 2 I. VAN RENSBURG SJ. Epidemiologic and dietary evidence for a specific nutritional predisposition to esophageal cancer. JNCI 1981 ;67:243-25 I. 22. ZaEGLER RG, MORRIS LE, BLOT WJ, et al. Esophageal cancer among black men in Washington DC. I[. Role of nutrition. JNCI 1981;67:1199- 1206. 23. TUYNS AJ, I~BOLI E, DOORNBOS G, et al. Diet and esophageal cancer in Calvados (France). Nutr Cancer 1987;9:81-92. 24. DECARL~ A, LI~'D P, NEGRI E, et al. Vitamin A and other dietary factors in the etiology of esophageal cancer. Nutr Cancer 1987;10:29-37. 25. Lu SH, OHSmMA H, FU HM, et al. Urinary excretion of N-nitrosamino acids and nitrate by inhabitants of high- and low-risk areas for esophageal cancer in northern China: endogenous formation of nitroso- proline and its inhibition by vitamin C. Cancer Res 1986;46:1485-1491. 26. YANG CS. Research on esophageal cancer in China: a review. Cancer Res 1980;40:2633-2644. 27. KALLNER AB, HARTMANN D, HORN1G n. On the requirements of ascorbic acid in man: steady-state turnover and body pool in smokers. Am J Clin Nutr 1981;34:1347-1355. 28. GA~RL~L GN, SCHRAGER TF, NEW~ERNE PM. Zinc deficiency, alco- hol, and a retinoid: association with esophageal cancer in rats. JNCI 1982;68:785-789. 29. BIRT DF. Update on the effects of vitamins A, C, and E and selenium on carcinogenesis. Proc Soc Exp Biol Med 1986;183:311-320. 30. McCORMICK DL, BAGG BJ, HULT~N DA. Comparative activity of dietary or topical exposure to three retinoids in the promotion of skin tumor induction in mice. Cancer Res 1987;47:5989-5993. 31. KLAASSEN CD. Absorption, distribution, and excretion of toxicants. In: Doull J, Klaassen CD, Amdur MO, eds. Casarett and Doull's toxicology: the basic science of poisons. New York: Macmillan, 1980:40. 32. MORTON JF. Plant products and occupational materials ingested by esophageal cancer victims in South Carolina. Q J Crude Drug Res 1973;13:2005-2022. 33. MORTON JF. The potential carcinogenicity of tea. J Environ Sci Health 1986;C4:203-223. 34. MORTON JF. Tannin and oesophageal cancer. Lancet 1987;2:327-328. 35. OTERDOOM HJ. Preventing oesophageal cancer. Lancet 1987;1:745. 36. MANCHANDE JP, METS T, KLION A. Aetiology of oesophageal cancer. Lancet 1987;1:1499. 37. VASSALLO A, CORREA P, DE STEFANI E, et al. Esophageal cancer in Uruguay. A case-control study. JNCI 1985;75:1005-1009. 38. SEG~ M. Tea-gruel as a possible factor for cancer of the esophagus. Gann 1975;66:199-202. 39. HtRAYAMA J. Diet and cancer. Nutr Cancer 1979;1:67-81. 40. KOLYCHEVA NI. Epidemiology of esophageal cancer in the USSR. In: Levin DL, ed. Cancer epidemiology in the USA & USSIL Wasb2ngton, DC: US Govt Print Off, 1980:191-198 [DHHS publication No. (NIH) 80-2044]. 41. MUNOZ N, V~C'rOPOA CG, CRESP~ M, et al. Hot mate drinking and precancerous lesions of the esophagus: an endoscopic survey in southern Brazil. Int J Cancer 1987;39:708-709. Vol. 80, No. 20, December 21, 1988 ARTICLES 1625