Tobacco Documents Online

II [CANCER RESEARCtl 48, 3282-3287. June f. 19RRI Smirking and Drinking in Rdation to Oral and Pharyngeal Cancer William J. Blot,' Joseph K. NIeLaughlin, Deborah NI. Winn, Donald F. Austin, Raymond S. Greenberg, Susan Preston-Martin, Leslie Bernstein, Janet B. Schoenberg, Annette Stemhagen, and Joseph F. Fraumeni, Jr. National Cancer Institute, Bethesda. Ataryland 20892 [~V. J. B., J. K. AL, J. F. F.]; National Center for Health Statistics, Hyattsville, ~laryland 20782 [D. AL W.J; California State Department of Health Services, Emeryville. California 94608 [D. F. .4.]; Emo~" University, Atlanta, Georgia 30322 JR. S. G.]; University of Southern California. Los Angeles. California 90033 iS. P-M., L. B.]; and .Vew Jersey Department of Itealth. Trenton. ,Vew Jersey 08625 [J. B. S., A. S.] ABSTRACT A case-control, study of oral'and pharyngeal' cancer conducted in four areas" of'the" united States provided' informhtior/ on the'tobacco and alcohol use of 1114 patients and 1268 population-ba~ed controls. Because of the large study size, it could be sho~n that the risks of these cancers among nondrinkers increased with amount smoked, and conversely that the risks among nonsmokers increased with the level of alcohol intake. Among consumers of both products, risks of oropharyngeal cancer tended to combine more in a multiplicative than additive fashion and were increased, mor.e than 35-fold among those w~ho consumed two or more packs ofclgai'e~te~ and more than four alcoholic d/'inks/day. Cigarette, cigar, and pipe smoking were separately implicated, .although it was shown for the first time that risk was not as high among male lifelong filter cigarette smokers. Cessation of smoking was associated with a sharply reduced risk of this cancer, with no excess detected among those having quit for 10 or more years, suggesting that smoking affects pri- marily a late stage in the process of oropharyngcal carcinogenesis. The risks varied by type of alcoholic beverage, being higher among those consuming hard liquor or beer than wine. The relative risk patterns were generally similar among whites and blacks, and among males and females, and showed little difference when oral and pharyngeal cancers were analyzed separately. From calculations of attributable risk, we estimate that tobacco smoking and alcohol drinking combine to account for ap- proximately three-fourths of all oral and pharyngeal cancers in the United States. INTRODUCTION Tobacco and alcohol are regarded as the major risk factors for oral and pharyngeal cancer in North America and in Euro- pean countries (1). It has been difficult to distinguish the separate effects of these agents, however, since drinkers of alcoholic beverages tend to be smokers, and vice versa. There- fore, large epidemiological investigations are required to eval- uate risks among persons exposed to only one of the two, precisely quantify the risk of one substance while adjusting for the other, and contrast effects according to the type of product and level of exposure. Herein is a report from such a study, a population-based case-control investigation of oropharyngeal cancer conducted over the past few years in four areas of the United States. METHODS ' " " " "" " " " Incident cases of pathologically confirmed primary oral and pharyn- geal cancer (International Classification of Diseases, 9th Revision; ICD 141-149), excluding cancers of the salivary gland (ICD 142) and nasopharyn, x (ICD 147), diagnosed during the period January 1, 1984 to March 31, 1985, were identified from the population-based cancer registries covering metropolitan Atlanta, Los Angeles, Santa Clara and San Mateo counties south of San Francisco-Oakland, and the state of New Jersey. All cases among white and black residents in these areas Received 11/16/87; revised 2/15/88; accepted 3/4/88. 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.~.C. Section 1734 solely to indicate this fact. ~To v. horn requests for reprints sh,mld be addressed, at Division of Cancer ['~tioh)~'. Epidemtt~h~gy and giostatistics Program. NItl, Bethesda. ~,ID 20892. between the ages of 18 and 79 were eligible and sought for interview after obtaining physician consent. .. Controls in each.area were selected from t.wo.pbpulation sources. • .Controls under age 65 3?ears ~ere identified by the random-digit-diali,~ technique for sampling households with telephones (2). Over 90% of the households in these areas are estimated to have telephones (3). For each residential phone number contacted, a household census was taken listing the age of occupants in the 18-64 years age range. Sufficient numbers of individuals were then selected so that the age (in 5-year groups), sex, and race (black, white) distribution of the controls was similar to .the ejpe.cted age-sex-race, distribution of the c.ases. The control selection process was sex specific, so that only a male or female (not both) was selected from any individual household. Controls age 65 years and over were systematically selected after a random start from rosters of residents in each area provided by the Health Care Financing Administration. Registry data on oropharyngeal cancer from prior years were used to determine the age (5-year groups), sex, and race distribu- tion of the controls. Interviews were sought with all eligible cases and controls and usually took place in their homes. For cases too ill, incapacitated, or deceased, interviews were sought with proxy respondents, first a spouse and then another first-degree relative if the spouse was unavailable. Trained interviewers used a structured questionnaire specifically designed fi:- this investigation. The questionnaire sought histories of tobacco and alcohol intake, dietary habits, residence, occupation, medical condi- tions, and other variables. The tobacco questions included screening questions to identify smokers as those who had smoked a total of I00 or more cigarettes or who had smoked cigars or pipes for 6 months or more, and drinkers as those who had drunk at least 20 beers, 20 drinks of wine, or 20 drinks of hard liquor over their entire lifetime. For those responding positively to a tobacco or alcohol screening question, a series of questions concerning age started use and usual amounts consumed were then asked. For alcohol consumption, usual intake ca weekends and weekdays was ascertained and then summed to obtain weekly consumption. The measure of association between oropharyngeal cancer risk and tobacco and/or alcohol intake used in this analysis is the odds ratio. Point estimates and 95% confidence intervals of summary. (adjusted) OR" were calculated using logistic regression analyses (4, 5). We also used logistic models to conduct formal tests of interaction between tobacco and alcohol. To determine whether the two agents combined more in an additive or multiplicative fashion, we fit models eontainbg a parameter representing a transition from a subadditive to a supra- multipticative combination of risks (6, 7). All analyses were adjusted • . for age, race, study. 19caJi'on,.and resl~ondent status (self versus n.ext-of- kin interview). Some analyses were adjuste~t also fbr education, and diet (quartile level of consumption of fruits and vegetables); however, these 2 variables showed only minimal confounding and adjustments for them left the OR estimates essentially unchanged. Adjustments for smoking and drinking used the categories shown in Table 4 (males) or 5 (females). Population attributable risk estimates were also calculated as measures of the proportion of all oropharyngeal cancers due ~o tobacco smoking and alcohol drinking (8). Excluded from the tables in the text are individuals with unknown values for the variable under study; hence the table totals are not always identical. We also conducted analyses using information-only from interviews from the cases and controls themselves (i.e., excluding next-of-kin interviews: 22% ofcases, 2% of controls), but results were bssentially the same and thus not presented in the text. : The abbrc~ iations used are: OR, odds ratio:. CI. confidence interval. 3282 RES~ Inte cases 75% interv age 6~' 73%a 76% v Talz contr( white i': origin years, Angel total). simila than c • • . .Few 'i have contr( arnon~ cases numb, yses ~ phone numb, analy~ Tob males, CO~TeS smok~ consu As adjust and r, smok~ little , eessat with i per durati seen i j and a~t Tabl t

SMOKING. DRINKING, AND OROPIIARYNGEAL CANCER RESULTS/ InterMews were completed for 1114 oropharyngeal cancer cases and 1268 controls; these numbers represent, respectively, 75% of all incident cases in the study areas and 76% of the rview-eligible controls. The response rate for controls under years was 79% at the household screening phase and 73% at .the field interview phase. Of controls 65 years and older, 76% were interviewed. Table 1 shows the distribution of the interviewed cases and controls by sex, race, and age. Most cases were male (68%) and white (83%). Among the white patients, 5% were of Hispanic origin. The median age at diagnos'is of the patients was 63 years, with 61% age 60 years or over. New Jersey and LOs A:'~geles contributed the largest numbers of cases (82% of the total). The percentages of controls by sex, race, and age were similar, although a somewhat higher proportion of controls than cases was selected in the West Coast study locations. Few (4%) cases among whites under age 65 years did not have a telephone in their households (a selection criterioh for controls under age 65 years), but the corresponding figure among black cases was 16%. Only 2% or less of white'and black cases age 65 years and older did not have Social Security numbers (a requirement for controls in this age group). Anal- yses were conducted deleting those under age 65 years without phones and th6se age 65 years and over without Social Security numbers, but results were essentially unchanged so only the analyses on the total sample are presented below. Tobacco. Nearly all the oropharyngeal cancer patients (93% males, 85% females) had been tobacco smokers, whereas the corresponding percentages for controls were significantly less (78% males, 53% females). Among the patients who were smokers, nearly all (95% males, 91% females) had also been tamers of alcoholic beverages. shown in Table 2, the risks of oropharyngeal cancer, adjusted for alcohol consumption (and age, race, study location, and respondent status), rose with the number of cigarettes smoked per day and the duration of cigarette smoking, showed little trend with age started smoking, and declined following cessation of smoking. Although not shown, the OR associated with increasing intensity (number of cigarettes usually smoked per day over the individual's smoking lifetime) rose within each duration category, and positive trends with duration were also seen in each intensity category. Although cessation, duration, ahd age started are colinearly related in this age-matched study Table I Numbers of interviewed oropharyngeal cancer patlents and controls according to sex, race, age, and study location Indicator No. 9f cases (%) No. of controls (%) Sex Male '762 (6.8) Female " 352 (32) Race VChite 020 (83) /~laCK 1'~4 (1 tJ Age <50 147 (13) 50-59 288 (26) 60-69 414 (37) 70+ 265 (24) Location New Jersey 492 (44) Atlanta 135 (12) Los Angeles 420 (38) San Francisco 67 (6) Total I 114 837 (66) " 431 (.3.4) 1065(84) 189 (15) 333 (26) 463 (37) 283 (22) 464(37) 138 (11) 525 (41) t4l (11) 1268 (with long-term quitters necessarily having shorter durations of smoking), the decline in risk following cessation was particu- larly rapid and marked. In contrast to an OR of 3 to 5 among current smokers, little or no elevation in risk was found among those who had quit smoking cigarettes for 10 or more years; for those who had quit smoking for 20+ years, the OR was actually below 1.0, although the confidence intervals at this level of cessation were wide. The strong trends of declining risk with cessation persisted after adjusting for amount smoked per day. Increased risks of oropharyngeal cancer were found regard- le~s of type of cigarette smoked, akhougk among.males the excess.was small for lifelongfilter smokers. After .adjust{ng for age, duration and usual amount smoked, and alcohol consump- tion, males who smoked only filter cigarettes (11%' of all cigarette smokers among controls) experienced 50% (95% CI = 30-80%) of the risk of smokers of only nonfilter cigarettes (34% of all smokers), while mixed filter and nonfilter smokers exp.eri.enced 80% (95% CI = 60-140%) of the risk compargd to pure nonfilier smokers. Little or"no reduciion i'n risk a~so- ciated with filter smoking was observed among females: after adjusting for duration and usual amount smoked, filter and mixed smokers, respectively, experienced 120% (95% CI = 60- 280%) and 90% (95% CI = 40-180%) the risk of nonfilter smokers. Sufficient numbers of men who had smoked cigars and/or pipes but not cigarettes were available for analysis of risks associated with these products. Oropharyngeal cancer risks was significantly elevated among those exclusively smoking cigars and/or pipes (OR = 1.9, 95% CI = 1.1-3.4, after adjusting for alcohol intake), with a positive trend associated with increasing numbers of cigars/pipes smoked. After adjusting for alcohol consumption among these noncigarette smokers, the OR rose to 16.7 (95% CI = 3.7-76.7) for men who smoked 40 or more cigars/week (14 cases, 1 control), and to 3.1 (95% CI = 1.1- 8.7) for those consuming 40+ pipefuls/week (12 cases, 7 con- trols.). Six % of cases and 7% of controls among males had used smokeless tobacco (primarily chewing tobacco), but nearly all were also smokers. Smokeless tobacco use was less common among females (3% cases, 1% controls), but users (primarily of snuff) generally were nonsmokers. Among nonsmoking fe- males, the OR was 6.2 (95% CI = 1.9-19.8) for users of smokeless tobacco (6 cases, 4 controls). All 6 cancers occurred in the oral cavity. Alcohol. Nearly all the cancer cases (94% males, 82% females) had been drinkers of alcoholic beverages, with significantly • -lower percentages among the controls (83% males, 60% fe- males). Among the cases who consumed alcoholic beverages, nearly all (95% males, 94% females) had .also smoked tbba.'~co.. Table 3 shows that the OR for.oropharyngeal cancer, cob-" trolled for smoking, rose with increasing total alcohol con- sumotion and with increasin~ intake of hard llouor and beer. Over one-half of the mate cases were heavy drinkers (30 or more drinks/week), compared to only 14% of the male controls. About one-fourth of the female patients drank this heavily, compared to 2% of the female controls. The smoking-adjusted excess risk associated with high consumption among both men and women was approximately 9-fold. The risk was relatively small among both men and women when alcohol intake was moderate (1 - 14 drin ks/week). The risk of oropharyngeal cancer differed according to the type of alcoholic beverage consumed. The trends were strongest for beer and hard liquor consumption and persisted after the 3283

SMOKING, DRINKING, AND OROPHARYNGEAL CANCER Table 2 Odds ratios for oropharyngeal cancer associated with smoking Exposure index Mal~s Females No. of No. of No. of No. of cases controls ORa 95,% CI cases controls ORa 95% CI ~Never smoked Cigar or pipe only Cigarette smoker No.' of cigarettes/day 1-19 20-39 40+ Years of cigarette smoking 1-19 20-39 40+. Age started smoking <17 17-24 25+ Years since s.topped gmoking cigarettes. 0 (never quit) I-9 10-19 20+ 50 185 1.0 54 202 1.0 52 56 1.9 1.1-3.4 0 0 659 593 1.9 1.3-2.9 298 229 3.0 80 173 1.2 0.7-1.8 60 312 288 2.1 1.4-3. I 145 262 130 2.8 1.8-4.4 93 45 13~ "0.8 - 0.5-I~ 15 286 281 1.9 1.2-2.8 127 313 171 3.6 2.3-5.6 153 2.0-4.5 104 1.8 1. I-2.9 94 3.6 2.3-5.8 31 6.2 3.6-11.3 -59. . 1.b. 0.5-i.9. 105 2.9 1.8-4.6 64 5.0 3.0-8.3 325 258 2.1 1.4-3.2 89 59 2.9 1.7-4.9 279 285 1.8 1.2-2.7 153 116 3.1 2.0--4.9 38. 47 1.8 0.9-3.3 54 54 2.8 1.6-4.8 129 4.7 3.027.3.. 39 1.8 0.9-3.6 35 0.8 0.4--1.9 26 0.4 0.1-1.4 • 4~S 239 3.4 ~3SS.I 258 64 98 1.1 0.7-1.9 24 56 114 1.1 0.7-1.9 10 43 " 141 0.7 0.4-1.2 4 All OR adjusted for alcohol consumption, age, race, study location, and respondent status. Table 3 Odds ratios for oropharyngeal cancer associated with drinking Females Males" Type of No. of No. of No. of No. of No. of alcohol drinks/wk cases controls OR* 95% CI cases controls ORa 95% CI All <i 40 139 1.0 63 171 1.0 1-4 71 206 1.2 0.7-2.0 75 129 1.2 0.7-1.9 5-14 99 219 1.7 1.0-2.7 72 93 1.3 0.8-2.1 15-29 154 150 3.3 2.0-5.4 55 29 2.3 1.2-4.5 30+ 389 118 8.8 5.4-14.3 87 9 9.1 3.9-21.0 Hard liquor <1 173 337 1.0b 135 278 1.0b I-4 134 240 1.0 0.7- 1.3 78 95 1.3 0.9-2. I 5-14 138 165 1.3 0.9-1.8 65 50 1.5 0.9-2.5 15-29 118 56 2.6 1.7-3.9 32 5 4.9 1.6-14.3 30+ 179 31 5.5 3.4-9.1 41 3 7.8 2.1-29.2 Beer <1 146 333 1.0# 180 343 1.0~ 1-4 130 231 1.2 0.8-1.7 73 60 2.2 1.4-3.6 5-14 141 161 1.7 1.2-2.4 48 20 2.9 1.5-5.6 15-29 134 " 62 3.4 2.7-5.1 24 7 2.3 0.9-6.5 30+ 195 44 4.7 3.0-7.3 27 1 18.0 2.1-159 Wine <I 497 490 1.0a 230 273 1.0~ 1-4 114 205 0.7 0.5-1.0 60 109 0.6 0.4-1.0 5-14 70 110 0.7 0.4-1.0 41 41 0.8 0.4-1.4 15-29 31 21 0.9 0.5-1.8 7 7 0.5 0,1-2.3 • 30+ 35 6 2.5 0.9-65 13 1 1.6 0.2-13.6 * OR adjusted for smoking, age, race, study location, and respondent status. n OR also adjusted for intake of beer and wine. NonsI Short 1-19/ 20-3¢ 40+/, Pipe/ Total' 95% t np signi = 32 stror W rood the c rive 0 risks The to th hear high cons # OR also adjusted for intake of hard liquor and wine. a OR also adjusted for intake of hard liquor and beer. "adjustment of ~ne fo~" th~ o~her.'There was litiie or a~ e~cess risk for wine drinking, reported by about one-third of both male except when consumption exceeded 4 drinks/day. There were few clear or consistent trends in risk with either age started drinking or duration of drinking hard liquor, beer, or wine. Differences in risk, e.g., between those starting con- sumption before age 20 years compared to age 30 years or after were typically on the order of 20% or less. There was evidence of a linear trend of rising risk with increasing years of beer drinking, but only among males, with little gradient with du- ration of wine or liquor drinking among either sex. The ques- tionnaire did not obtain information necessary to calculate an OR associated with cessation of drinking. Smoidng ~hd Drinking' interreiations. Tables 4 (male~) and.5 (females) ~how that the increasing trends in risk with smoking Conversely, the increasing trends with alcohol consumption were seen across all smoking categories, except for female nonsmokers, among whom few women (either cases or contre*5) were heavy drinkers. In combination, heavy smoking and drin x- ing resulted in very large increases in oropharyngeal cancer risk, with an OR of 38 among males and exceeding 100 among females. We examined interaction (on a multiplicative scale) by fitting logistic regression models containing the smoking and alcohol categorical variables of Tables 4 and 5 as well as their cross- products. Global multiple degree of freedom tests uncovered a 3284 tion- md/ latio and ' amo amo to h, 200 ages patt¢ simi smo) ages forb ,',ca',

SMOKING, DRINKING, AND OROPHARYNGEAL CANCER • Table 4 Odds ratios for oropharyngeal cancer among males according to amount of smoking and drinking All OR adjusted for race, age, study location, and resp6ndent status. " -95~ CI No. of alcoholic drinks/wk Smoking status <1 1--4 5-14 15-29 30+ Total" I,,~Nonsmoker 1.0 (12, 66)t' 1.3 (12, 52) 1.6 (15, 39) 1.4 (5.21) 5.8 (6, 7) ~qhort duration or former* 0.7 (8, 42) ~., (24, 61) 1.4 (21, 90) 3.2 (25, 49) 6.4 (43, 37) ll~-19/day for 20+ yr 1.7 (2, 6) 1.5 (7, 21) 2.7 (8, 18) 5.4 (16, 18) 7.9 (22, 14) 20-39/day for 20+ yr 1.9 (8, 17) 2.4 (17, 34) 4.4 (28, 40) 7.2 (52, 42) 23.8 (145, 33) 40+/day for 20+ yr 7.4 (9, 4) 0.7 (6, 14) 4.4 (19, 19) 20.2 (43, I 1) 37.7 (148, 21) Pipe/cigar only 0.6 (I, 4) 1.0 (5, 24) 3.7 (8, 13) 4.7 (13, 9) 23.0 (25, 6) Totala 1.0 (40, 139) 1.2 (71,206) 1.7 (99, 219) 3.3 (154, 150) 8.8 (389, 118) 95% CI 0.7-2.0 1.0-2.7 2.0-5.4 5.4-14.3 1.0 (50, 185) 1.1 (121,279) 1.6 (55, 77) 2.8 (250, 166) 4.4 (225, 69) 1.9 (52, 56) 0.7-1.7 0.9-2.7 1.8-4.3 2.7-7.2 1.1-3.4 OR adjusted for drinking and relative "to nonsmokers. Parentheses. numbers of cases and controls. Quit smoking cigarettes for 10 or more years or smoked for less than 20 years. Over 85% of these individuals were ex-smokers for 10+ years. OR adjusted for smoking and relative to those who drank less than 1 alcoholic drink/week. Table 5 Odds ratios for oropharyngeal cancer among females according to amount of smoking and drinking All OR adjusted for race; age, study location, and respondent status. . .No. of alcoholic drin "ks/wk . , Smoking status <1 I--4 5-14 15-29 30+ Total" "~5% CI Nonsmoker 1.0 (36, 112)b 0.7 (11,'62) . 1.3 (7, 23) 0.0 (0, 3) 0.0 (0, 2) 1.0 (54, 202) Short duration or formere 1.0 (7, 27) 1.6 (8, 21) 0.4 (4, 30) 1.1 (3, 10) -(3, 0) 1.0 (25, 88) 0.5-1.8 1-19/day for 20+ yr 0.9 (4, 13) 5.1 (22, 15) 2.8 (11, 15) 4.6 (3, 3) 11.0 (9, 3) 3.0 (49, 49) 1.9-5.2 20-39/day for 20+ yr 2.2 (12, 19) 2.7 (20, 25) 6.9 (35, 18) 12.4 (31, 9) 46.0 (38, 3) 4.4 (136, 74) 2.7-7.2 40+/day for 20+ yr -(4, 0) 9.3 (14, 6) 7.8 (15, 7) 18.0 (18, 4) 107.9 (37, I) 10.2 (88, 18) 5.2-20.4 Total't 1.0 (63, 171) 1.2 (75, 129) 1.3 (72, 93) 2.3 (55, 29) 9.1 (87, 9) 95% CI 0.7-1.9 0.8-2.1 1.2-4.5 3.9-21.0 OR adjusted for drinking and relative to nonsmokers. Parentheses. numbers of cases and controls. Quit smoking cigarettes for 10 or more years or smoked for less than 20 years. OR adjusted for smoking and relative to those who drank less than 1 alcoholic Approximately two-thirds of these individuals were ex-smokers for 10+ years. drink/week. Qsignificant interaction for males (~2 with 20 degrees of freedom 32, P = 0.05) but not females, due mainly to an exceptionally strong effect of alcohol among pipe and cigar smokers. We also contrasted additive versus multiplicative relative risk models. The additive models among both males and females fit the data rather poorly and could be discarded. The multiplica- tive effect (and departure from additivity) is exemplified by the risks associated with the highest levels of exposure in Table 4. The OR of 37.7 among heavy smokers-heavy drinkers is close to the multiple of the OR values of 5.8 and 7.4 associated with heavy exposure to one substance in the absence of the other. Attributable Risks. Since smoking and drinking habits are • highly correlated, with few abstainers of one product among consumers of the other, we calculated estimates of the popula- tion-attributable risk of oropharyngeal cancer due to smoking and/or drinking rather than due to each separately. The popu- lation-attributable risks ¢vere 80% for males, 61% for females, and 74% overall. When we partitioned the sample according to amount of tobacco and al6ohol consumed, it was found that among the smoking- and alcohol-related tumors (representing 74% of all cases) approximately two-thirds could be attributed to heavy consumption (t.e., smoking 2 or more pacKs/Oay /or 20 or more years and/or drinking 30 or more alcoholic bever- ages/week)• Racial Differences. Although not shown in the tables, the patterns of risk among drinkers and smokers were generally similar for blacks and whites. The prevalence of drinking and smoking, however, differed. Among male controls, the percent- Aages of moderate to heavy drinkers and/or smokers were higher Qr blacks (73%) than whites (55%)• Much of the racial discrep- ancy was accounted for by a considerably higher prevalence of :~cavy drinking among the black (22%) than white (I 3%) general population, with smoking differentials not as great. The higher prevalence of exposure among blacks was not observed among females, of whom only 2% of the black and 7% of the white controls were heavy smokers or drinkers. Smoking and drinking were also less prevalent and less strongly related to oropharyn- geal cancer risk among Hispanics, but data were sparse. Anatomic Sites. Calculations were conducted separately for tongue cancer (ICD 141), other oral cancer (ICD 143-145), and pharyngeal cancer (ICD 146, 148, and 149), comprising, respectively, 28%, 38%, and 34% of the cancer cases. As shown in Table 6, smoking (after adjusting for alcohol intake) and drinking (after adjusting for smoking) were significantly related to each site of cancer. The trends with smoking and drinking were slightly weaker for tongue cancer than other sites among men. Among females the effects of smoking and drinking tended to be stronger for pharyngeal cancer than oral cancer. Pipe and cigar smoking was more closely associated with other oral cancer (including cancers of the floor of the mouth and buccal mucosa) than either tongue or.pharyngeal c'~incer. Tobacco and alcohol have long been implicated as risk factors for oropharyngeal cancer. Reports early in this century de- scribed smoking and drinking as unusually common among oral cancer patients (9). It was not until the late 1950s and early 1960s, however, that epidemiological studies began to depict the association in quantitative terms. The initial cohort studies of smokers showed 4-fold or greater oropharyngeal cancer death rates among smokers compared to nonsmokers (10-12). Case-control studies in several areas of the United States (13-15) documented an elevated risk of oral cancer 3285

SMOKING. DRINKING, AND OROPHARYNGEAL CANCER Table 6 Odds ratias for oropharyngeal cancer subsites associated with smoking and drinking by sex Males Females Other Other Tongue oral Pharynx Tongue oral Pharynx Amount smokeda Nonsmoker 1.0 1.0 1.0 1,0 1.0 1.0 Light or former " 0.8 1.2 1.5 1.3 0.8 1.6 1-19/day for 20+ 1.4 1.8 1.5 2.6 2.0 9.7 yr 20-39/day for 1.9 3.8 3.4 2.8 4.6 14.3 20+ .~T 40+/'day for 20+ 3.2 5.2 5.8 8.1 9.7 36.7 yr Pipe/cigar only 1.9 '2.7 1..6 No. of alcoholic drinks/wkn < 1 1.0 1.0 1.0 1.0 1.0 1.0 1-4 0.7 1.8 1.4 1.2 1.0 ] .6 5-14 t.3 2.4 1.7 1.I 1.5 1.4 15-29 2.4 4.3 4.3 2.0 2.8 3.5 30+ 6.0 12.3 10.4 11.0 6.6 15.0 Total no. of cases by 216 272 273 104 151 100 site "OR adjusted for race, age, study location, responden;t status, and drinking. '~ OR adju_sted for race, age, study location, respondent statt~s, and smoking. associated with drinking as well as smoking, although adjust- ment "for one factor by the other was generally not made. Although recent studies have confirmed that alcc~hol and tobacco are important causes of oropharyngeal cancer (16), questions remain about the relative impact of drinking com- pared to smoking, the effects of different types of alcohol and methods of smoking, and the interrelationship between these risk factors. The data presented in this report, derived from the largest population-based case-control investigation of oropha- ryngeal cancer, help resolve these etiological issues. We found that tobacco smoking and alcohol drinking sepa- rately increase the risk of oral and pharyngeal cancer. Strong dose-response relationships were observed for each substance after tightly controlling for exposure to the other. Noteworthy were trends of increasing risk with smoking among nondrinkers and with drinking among both nonsmokers and short duration and former smokers. Although similar effects have been re- ported previously (17-19), the larger study size available to us provides reliable confirmation of the excess risks associated with one substance in the absence of the other. Because of the population-based nature of the study, we could calculate population-attributable risk estimates. In total we estimate that approximately three-fourths of all oral and pha- ryngeal cancers in the study areas, and probably in the United States, are caused-by smoking and drinking, with most of the cancers due to heavy consumption. We did not attempt to assign attributable risk percentages separately to alcohol and tobacco since nearly all male patients and most female patients used both products. Among males the adjusted OR values were somewhat higher than for moderate (2.8) and heavy (4.4~ smoking, as were the prevalences of these habits in the general population (32% of the controls were moderate/heavy drinkers and 25% moderate/heavy smokers), so that drinking seems to contribute more than smoking to oropharyngeal cancer risks. However, among females the effect of smoking appeared some- what more pronounced than drinking. Some investigators (I 7, 19) have noted roughly comparable effects, while others have described alcohol (18, 20-22) or tobacco (23) as the more important risk factor. We found an increased risk of oropharyngeal cancer among both filter and nonfilter smokers. Among males, however, life- long filter smokers experienced only one-half the risk of lifelong n0nfilter smokers. Filter smoking'also has beefi associated with a reduced risk of lung cancer,S4), but to our knowledge this is the first, report of a lowered risk of oropharyngeal cancer. Smoking only pipes G~:-cigars increased risk hearly as much as cigarettes, particularly when intake was heav3". Hence it appears that any smoked tobacco can cause these cancers. Smokeless tobacco is also a strong risk factor for oral cancer (25). The prevalence of snuff dipping or tobacco chewing in the study population was low, with nearly all male users of smokeless tobacco a/so being smokers, but a significant excess ora~ cancer risk was detected among nonsmoking women who used snuff. consistent with findings of other studies (25). The sharp declines in risk of oropharyngeal cancer following cessation of smoking were remarkable. Indeed, after 10 years both male and female quitters experiences no increase in risk relative to nonsmokers. The drop in risk in such a relatively short time suggests that smoking primarily affects the late stages of oral carcinogenesis. A reduction in risk of oral cancer following cessation has been observed elsewhere (19), and a rapid decline among ex-smokers was also reported recently for bladder cancer (26). Similar findings for lung cancer (16) un- derscore the importance of smoking cessation as a means of preventing various types of cancer. Among long-term quitters, however, there was still a marked gradient in risk associated with alcohol intake, so that reduction in use of both products appears required to approach the low baseline risk of oropha- ryngeal cancer among abstainers of tobacco and alcohol. Increased risks of oropharyngeaI cancer were associated with hea~3~ consumption of each type of alcohol (hard liquor, beer, and wine), suggesting a role for ethanol, the common ingredient in these beverages. Experimental studies have not implicated alcohol itself as a carcinogen, although it may promote carci- nogenesis by a variety of mechanisms, including nutritional deficiencies associated with heavy drinking, the effects of con- taminants and congeners in alcoholic beverages, the induction of microsomal enzymes that enhance the metabolic activation of tobacco or other carcinogens, and the capacity of alcohol to solubilize carcinogens or enhance their penetration in oropha- ryngeal tissues (27, 28). Our questionnaire did not obtain information on age at cessation of drinking, so we could not calculate risks for current versus former drinkers to directly evaluate the promotional effect of recent alcohol exposure, but the absence of clear trends with age started or duration of drinking argues in favor of a late- rather than early-stage effect of alcohol. We did find that drinking was interrelated with smoking, one enhancing the effect of the other, with the excess risk greatest among heavy consumers of both products; how- ever, we also detected a deleterious effect of heavy alcohol consumption among nonsmokers, indicating that tobacco is not a requisite cofactor for alcohol-related cancer. Further studies ~re needed to determine whether alcohol promotes the effects of other exposures, such as human papillomavlruses (~9), tna~ remain to be clarified as risk factors for oropharyngeal cancer. In addition, heavy intake of alcoholic beverages is correlated with nutrient deficiency (30), which appears to contribute in. dependently to oral carcinogenesis (31-33). Although heavy consumers of all types of alcohol were at increased risk of oropharyngeal cancer, the risks were greatest for hard liquor and beer and least for wine drinkers. Reasons for the differential in risks are not clear, but it is possible that ingredients in alcoholic beverages other than ethanol are in- volved, for example, nitrosamines and polycyclic hydrocarbons 3286

SMOKING. DRINKING. AND ORC~PH ~.RY'~GFAL C.'~('FR in beer and whiskey (34). In our stud)', alter adjusting for smoking and for consumption of liquor and beer, there was actually some deficit in risk among light and moderate wine drinkers. This may be due to factors associated with wine prinking, although adjustments for education (i.e., social class) and diet still did not eliminate the lowered risk. It is noteworthy that the dose-responSe relationship for wine consumption fol- lowed a cur~ilinear form, with no increases at up to 2 drinks/ day, but sharply rising risks thereafter. In contrast, risks rose nearly linearly with increasing consumption of beer and liquor, and were elevated at nearly all levels of intake. National incidence and mortality date 'in the United States :;how higher rates of oropharyngeal cancer among blacks than ;vhites, particularly for males (35). Applying the risk estimates derived herein to the percentages of drinkers observed among the controls in our population-based study, the greater preva- lence of drinking among black males would account for about one-half of their 40-50% higher rates of oropharyngeal cancer compared to white males in the 1970s. Reasons for the remain- ing excess among black men and the 20% higher rates among black than white women will be explored in a sephrate report. Similar conditions indicate that most of the more than 2-fold higher incidence of oropharyngeal cancer among males is due to higher levels of tobacco and alcohol consumption compared to females. " In summa_D', this large population-based case-control study implicates tobacco smoking and alcohol drinking as the major determinants of oral and pharyngeal cancer. The strong dose- response relationships leave little doubt that either product .'.done can induce these cancers, although most cases result from the combined effect of smoking and drinking. Thus, measures to prevent oral and pharyngeal cancer should be aimed at ~reducing intake of both alcohol and tobacco. The study also I'~'provides new leads regarding differential effects by type of product, including the lower risks associated with filter ciga- rettes and with intake of wine compared to beer or liquor. Furthermore, etiological mechanisms were suggested by a sharply reduced risk following smoking cessation, indicating that smoking, like drinking, probably affects a late stage in oral carcinogenesis. ACKNOWLEDGMENTS We thank Dr. Jay Lubin for statistical advice and computation, David Main for computer support, Louise Hanson for field supervision, and Susan Privot for manuscript preparation. REFERENCES 1. tMahboubi, E., and Sayed, G. M. Oral cavity and pharynx. In: D. Schottenfeld and dr. F. Fraumeni, Jr., (eds.), Cancer Epidemiology and Prevention, pp. 583-595~Y'hiladelphia: W. B. Saunders Co., 1982. 2. Hartge, P., Brinton, L., Rosenthal, J., Cahill, J., Hoover, R. N., and Waks- 34. Am. J. Epidemiol., 120: 825-833, 1984. 3. Bleinfante. A. An analysis of telephone penetration. Washington, DC: Fed- 35. eral Communications Commission, 1986. 4. Breslow, N. E., and Day. N. E. Statistical Methods in Cancer Research. Analysis of Case-Control Studies. Lyons, France: International Agency for Research on Cancer, 1980. 5. Harrell, F. E., Jr. The Logist Procedure. Supplemental Users Guide. Gary, NC: SAS Institute, t983. 6. Thomas, D. C. General relative risk models for survival time and matched case-control analysis. Biometrics. 37: 673-686, 1981. 7. Lubin. J. tt., and Gaffe~,, W. Relative risk models in assessing the joint effects of multiple factors. Am. J. Indust. Med., 13: 149-168, 1988. 8. Whittemore, A. S. Estimating attributable risk from case-control studies. Am. J. Epidemiol., 117: 76-85, 1983. 9. Abbe, R. Cancer of the mouth. NY Med. J., 102: 1-2, 1915. 10. Hammond. E. C. Smoking in relation to the death rates of 1 million men and women. Natl. Cancer Inst. Monogr., 19: 127-204, 1966. 11. 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. 12. Doll, R.. and Peto, R. Mortality in relation to smoking: 20 years' observation in male British doctors. Br. Med. J., 2: 1325-1536, 1976. 13. W.vnder, E. L, Bross, I. J., and Feldman, R. M. A study of the etiological factors in cancer of the mouth. Cancer (Phila.), 19: 1300-1323, 1957. 14. Vogler, W. R., Lloyd, J. W., and ,Milmore, B. K. A retrospective study of etiological factors in cancer of the mouth, phar2,'nx, and larynx. Science (Wash. DC), 15: 246-258, 1962. 15. Vincent, R. G., and Marchetta, F. The relationship of the use of tobacco and alcohol to cancer of the oral cavity, pharynx, or larynx. Am. J. Surg., 106: 501-505, 1963. 16. Surgeon General. The Health Consequences of Smoking: Cancer. Washing- ton. DC: Department of Health and Human Services, 1982. 17. Rothman, K., and Keller, A. The effect of joint exposure to alcohol and tobacco on risk of cancer of the mouth and pharynx. J. Chronic Dis., 25: 711-716, 1972. 18. Elwood, J. M., Pearson, J. C. G., Skippen, D. H., and Jackson, S. M. Alcohol, smoking, social and occupational factors in the etiology of cancer of the oral cavity, pharynx, and larynx. Int. J. Cancer, 34: 603-612, 1984. 19. Wynder, E. L., and Stellman, S. D. Comparative epidemiology of tobacco- related cancers. Cancer Res., 37: 4608-4622, 1977. 20. Mashberg, A., Garfinkel, L., and Harris. S. Alcohol as a primary, risk factor in oral squamous carcinoma. CA Cancer J. Clin., 31: 146-156, 1981. 21. Brugere, J., Guenel, P., Lecterc, A., and Rodriquez, J. Differential effects of tobacco and alcohol in cancer of the lar3'nx, pharynx, and mouth. Cancer (Phila.), 57: 391-395, 1986. 22. Martinez, I. Factors associated with cancer of the esophagus, mouth, and pharynx in Puerto Rico. J. Natl. Cancer Inst., 42: 1069-1094, 1969. 23. Graham, S., Dayal, H., Rohrer, J., Swanson, M., Sultz, H., Shedd, D., and Fischman, S: Dentition, diet, tobacco, and alcohol in the epidemiology of oral cancer. J. Natl. Cancer Inst., 59:1611-1617, 1977. 24. Lubin, J. H., Blot, W. J., Berrino, F., Flamant, R., Gillis, C. R., Kunze, M., Schmahl, D., and Visco, G. Patterns of lung cancer risk according to type of cigarette smoked. Int. J. Cancer, 33: 569-576, 1984. 25. Surgeon General. The Health Consequences of Using Smokeless Tobacco. Washington DC: Department of Health and Human Services, 1986. 26. Hartge, P., Silverman, D., Hoover, R., et aL Changing cigarette habits and bladder cancer risk: a case-control study. J. Natl. Cancer Inst., 78: 1119- 1125. 1987. 27. Lieber, C. S., Seitz, H. K., Garro, A. J., and Warner, T. M. Alcohol-related diseases and carcinogenesis. Cancer Res., 39: 2844-2850, 1979. 28. Tuyns, A. J. Alcohol. In: D, Schottenfeld and J. F. Fraumeni, Jr., (eds.), Cancer Epidemiology and Prevention, pp. 293-303. Philadelphia: W. B. Saunders Co., 1982. 29. DeVilliers, E., Weidauer, H., Otto, H., and ZurHausen, H. Papillomavirus DNA in human tongue carcinomas. Int. J. Cancer, 36: 575-578, 1985. 30. Ziegler, R. G. Alcohol-nutrient interactions in cancer etiology. Cancer (Phila.), 58." 1942-1948, 1986. 31. Marshall, J., Graham, S., Mettlin, C., Shedd, D., and Swanson, M. Diet in the epidemiology of oral cancer. Nutr. Cancer, 3: 145-149, 1982. 32. Winn, D. M., Ziegler, R. G., Pickle, L, W., Gridley, G., Blot, W. J., and Hoover, R. W. Diet in the etiology of oral and pharyngeal cancer among women from the southern United States. Cancer Res., 44:1216-1223, 1984. 33. Wynder, E. L., Hultberg, S., Jacobsson, F., and Bross, I. J. Environmental factors in cancer of the upper alimentary tract: Swedish study with special reference to Plummer-Vinson (Paterson-Kelly) syndrome. Cancer (Phila.), 10: 470-487, 1957. Walker, E. A., Castegnaro. M., Garren, L., Toussaint, G., and Kowaiski, B. Inst., 63: 947-951, 1979. Young, J. L., Percy, C. L., and Asire, A. J. Surveillance, epidemiology, and end results: incidence and mortality data 1973-77. Natl. Cancer Inst. Mon- ogr., 57: 54-55, 1981. 3287