Squamous cell carcinoma of the upper aerodigestive tract. A case comparison analysis

Squamous Cell Carcinoma of the Upper Aerodigestive Tract

A Case Comparison Analysis

MARGARET R. SPITZ, MD, MPH,' JOHN J. FUEGER, MA,' HELMUTH GOEPFERT, MD.t WAUN KI HONG, MD,* AND GUY R. NEWELL, MD*

Although the etiologic importance of tobacco in risk of upper aerodigestive malignancies is unques- tioned, quantification of subsite-specific risks is less well delineated. Risk estimates from this case- control study are derived from self-administered comprehensive risk factor questionnaires distributed to newly registered patients at The University of Texas M.D. Anderson Hospital and Tumor Institute, Houston. Cases included 185 white patients with histologically confirmed squamous cell carcinoma of the upper aeradigestive tract. An equal number of age-frequency and sex-frequency matched patients was randomly selected from the same patient population excluding only patients with diagnoses of squamous cell carcinoma of any site. A statistically significant dose-response relationship for three categories of cigarette pack-years was evident for both males (odds ratios [ORJ = 1.8,4.0, and 7.5) and females (OR = 1.5,9.0, and 12.0). Highest risks were documented for laryngeal cancer (OR = 15.1) and lingual cancer (OR = 14.5). There was interaction between alcohol use and smoking among men, but no independent effect of alcohol consumption among either gender. After 15 years of smoking abstinence, males no longer exhibited increased risk (OR = 1.0) whereas the risk for females after 15 years of cessation was 1.5. There were also significantly increased risks among men associated with snuff dipping, cigar, and pipe use (OR = 3.4,2.8, and 1.8, respectively). The differences in the magnitude of the risk estimates and dose-response curves by subsite and by sex suggest a variable susceptibility to carcinogenic action.

Cancer 61:203-208,1988.

ESPITE THEIR anatomic contiguity, head and neck D cancers are a disparate and morphologically di- verse group. Although risk factors for these cancers as a group have been fairly extensively studied, quantification of risk for specific sites is lacking, particularly with respect to tobacco exposures. Since not all sites within the head and neck area are equally susceptible to carcinogens, it is reasonable to assume that site-specific risks may differ. This study reports on quantification of risk factors for squamous cell carcinoma of selected sites within the head and neck in a hospital-based case-control analysis, with particular reference to tobacco and alcohol exposures.

From the *Department of Cancer Prevention and Control, ?Depart- ment of Head and Neck Surgery, and $Department of Medical Oncol- ogy, The University of Texas M. D. Anderson Hospital and Tumor Institute at Houston, Houston, Texas.

The authors thank E. Neely Atkinson, PhD, Division of Biomathe- matics, for statistical expertise and advice.

Address for reprints: Margaret R. Spitz, MD, Department of Cancer Prevention and Control, Box 189, The UT M.D. Anderson Hospital and Tumor Institute, 15 I5 Holcombe Boulevard, Houston, TX 77030.

Accepted for publication July 27, 1987.

Materials and Methods

As part of the registration procedure, all adult patients at The University of Texas M.D. Anderson Hospital and Tumor Institute, Houston, are asked to complete a detailed, self-administered risk factor questionnaire. Cases for this analysis included all US resident patients registered at this institution from January 1985 through Feb- ruary 1987 who completed and returned this questionnaire, and who presented with histologically confirmed squamous cell carcinoma of the larynx, tongue, orohypopharynx, and oral cavity. There were too few patients of other ethnicities meeting the study criteria to enable meaningful ethnic-specific comparisons. Also excluded from this investigation because of epidemiologic dissim- ilarities in etiology were patients with cancers of the skin, lip, salivary gland, esophagus, nasopharynx, or nasal cavity. We also excluded patients in whom the primary site of disease could not be determined with accuracy, or whose histologic diagnosis was not verified at this institution.

Controls were randomly selected from all other US resident white patients registered at the institution who had completed the questionnaire. Controls were fre-

203

204 CANCER January 1 1988 Vol. 61

TABLE I . Distribution of Primary Site of Malignancy Among Case Group

Men Women

Site No. Percent No. Percent

Larynx 50 38.2 15 27.8

Oroh ypopharynx 15 11.4 8 14.8 Floor of mouth 14 10.7 3 5.6 Other oral cavity* 27 20.6 16 29.6

Total 131 100.0 54 100.0

Tongue 25 19.1 12 22.2

* Gingiva (lo), buccal mucosa (6). retromolar trigone (8). palate ( I 3). oral cavity or pharynx not otherwise specified (6).

quency matched to the cases on age at registration (k5 years) and sex. Excluded as potential controls were patients with squamous cell carcinoma of any site (including all patients with bronchogenic carcinoma) because they would be most likely to share common exposures with the cases, thereby, biasing odds ratio estimates toward unity.

Besides eliciting information on life-style, diet, and medical, surgical, and family history, the survey includes sections on miscellaneous putative risk factors such as use of mouthwash and hair dye. Detailed information about smoking history and tobacco use is obtained. With respect to cigarette exposure, patients are requested to record the age at which they started smoking, the average daily number of cigarettes smoked, de- gree of inhalation and if they had ceased smoking, and the duration of abstinence. Some patients who respond affirmatively to smoking may fail to quantitate their smoking habits, so that pack-year data cannot be computed for all smokers. There are similar questions on pipe and cigar smoking, as well as use of smokeless tobacco products.

TABLE 2. Distribution of Diagnoses Among Control Patients

Men Women

Diagnosis No. Percent No. Percent

Malignant Melanoma Gastrointestinal tract Urinary system Prostate and testis Uterus Breast Bone Soft tissue Hematopoietic Brain

Nonmalignant.

Total

21 16.0 4 20 15.3 7 16 12.2 4 33 25.2

5 1 0.8 17 5 3.8 0 7 5.3 2

16 12.2 6 2 I .5 0

10 7.1 9

131 100.0 54

~

7.4 12.9 7.4

9.3 31.5

3.7 1 1 . 1

16.7

100.0

-

-

* Includes miscellaneous benign diseases of gastrointestinal tract, genitourinary tract, and central nervous system.

Subjects also are asked to classify their beer, wine, and hard liquor consumption into five ordinal categories ranging from “never” to “regular use” (several times weekly). The amount of alcohol consumed in terms of cans, glasses, or shots also is obtained. For ease of analysis, patients who never drank alcohol and those who reported infrequent use (a few times a year) were combined into a nonuser group.

Odds ratios were calculated to test for significant asso- ciations using Epilog software.’ The Mantel-Haenszel test for overall association in stratified analysis was used with computation of adjusted odds ratio estimates to achieve simultaneous control of more than one variable.2 Logistic regression utilizing an iterative maximum likelihood approach was accomplished by using the LO- GRESS interactive menu-driven p r ~ g r a m . ~

Results

One-hundred thirty-one men and 54 women met the case criteria for inclusion into the study. Ages ranged from 29 to 95 years for men (mean, 60.5 years) and 34 to 87 years for women (mean, 62.4 years). The distribution of primary sites, which did not differ significantly between men and women, is presented in Table 1. Lar- yngeal cancer was the most common diagnosis among both men (38.2%) and women (27.8%). Within this site category, involvement of the glottis occurred in 28 (56%) of the male laryngeal cancers compared with four (27%) for females. In contrast, the supraglottis wasdocu- mented as the primary site in 20% of male laryngeal cancers and 47% of females. Tongue cancers were the second most common primary site in both sexes. The miscellaneous “other oral cavity” category included ten patients with gingival lesions, six with buccal mucosa, and eight with retromolar trigone lesions; whereas 13 and six patients had cancers located in the palate and pharynx, respectively.

Among controls, all but ten men and nine women had a confirmed malignant diagnosis (Table 2). For men, these included 33 (25.2%) with prostatic and testicular cancers, 2 I ( 16%) with melanoma, and 20 ( 1 5.3%) with gastrointestinal (primarily colon) cancers. Breast cancer (31.5%) was the most common malignant diagnosis among women in the control group, followed by gastrointestinal malignancies ( 1 2.9%). Of the 20 control patients with urinary system cancers, 15 had bladder cancer.

As defined by the selection criteria, the age distribution of the cases and controls was identical. There were no significant differences between cases and controls with respect to religion, medical and surgical past history, previous radiotherapy, past history of cancer, wearing of dentures, or use of mouthwash.

No. I RISK FACTORS FOR UPPER AERODIGESTIVE CANCERS - Spirz et a/. 205

Data on cigarette smoking were available for all but four men in the case group and three women in the control group. The prevalence of smoking was 89% ( I 13 of 127) for males and 83% (45 of 54) for female cases compared with 64% (84 of 131) and 47% (24 of 51), respectively, for the control group. The overall odds ratios for cigarette smoking were 4.5 (2.4, 8.5) for men and 5.6 (2.4, 13.4) for women. The distribution of cases and controls according to three categories of computed pack-years of cigarette smoking is displayed in Table 3. Pack-year data were not available for an additional ten cases and six controls. Thus, computation of pack-year exposure was based only on 171 cases (121 men, 50 women) and 176 controls (1 27 men, 49 women). There was a consistent dose-response relationship for both men and women with higher risks evident for women at successive smoking strata. Among men, there was a linear increase in risk from 1.8 to 4.0 to 7.5. For women, the corresponding risks were 1.5, 9.0, and 12.0. In both instances, linear trend analysis was statistically significant.

Table 4 displays the odds ratios by pack-years smoked for five primary cancer sites among male patients. There were too few females to enable meaningful site-specific analyses. Highest risks across each pack-year stratum were noted for patients with laryngeal cancer. A dose- response effect was evident for these patients, with risk increasing from 2.7 for the lowest category of smoking

TABLE 3. Risk Estimates for All Study Sites by Cigarette Pack-Years

No. of No. of Odds 95% confidence Pack- years* cases controls ratio limits

Men Nonsmokers 14 47 1 .o -

1-24 12 23 1.8 0.7. 4.4 25-49 37 31 4.0 1.9, 8.5 so+ 58 26 7.5 3.7, 15.3

Trend P = <0.01

Women Nonsmokers 9 27 1 .o -

1-24 6 12 I .5 0.4, 5. I 25-49 15 5 9.0 2.1, 29.5 so+ 20 5 12.0 3.8, 38.0

Trend P = 10.01

Referent category was nonsmokers. Pack-year data were unavailable for six male and four female pa-

tients who smoked and four male and two female control smokers.

to 15.1 for the heaviest smokers (P < 0.0 1 for trend). For men with lingual cancer there was a two-fold excess risk in the first smoking stratum, which rose steeply to 14.5 among heaviest smokers. For patients with lesions of the floor of the mouth, there were seven-fold and nine-fold elevated risks in the two heaviest smoking categories. Lowest risk estimates were evident across each stratum for orohypopharyngeal and other oral cavity cancers with heaviest smokers exhibiting nonsignificant excess risks of 3.0 and 2.1, respectively.

TABLE 4. Site-specific Risk Estimates for Cigarette Pack-Years (Males Only)

No. of No. of Odds 95% confidence Primary site Pack-years cases controls ratio limits

Larynx Nonsmokers 3 47 I .o - 1-24 4 23 2.7 0.6, 12.6

25-49 15 31 7.6 2.3, 24.9 sot 25 26 15.1 5.0. 45.3

Trend P = <0.01 Tongue Nonsmokers 2 47 1 .o -

1-24 2 23 2.0 0.3, 14.9 25-49 4 31 3.0 0.6, 16.4 so+ 16 26 14.5 4.0, 52.5

Trend P = <0.01 Orohypopharynx Nonsmokers 3 47 1 .o -

1-24 I 23 0.7 0.1, 6.8 25-49 4 31 2.0 0.4. 9.4 so+ 5 26 3.0 0.7, 12.9

Trend P = 0.13 Floor of mouth Nonsmokers I 47 I .o -

1-24 1 23 2.0 0. I , 32.2 25-49 5 31 1.6 1 . 1 , 51.2 so+ 5 26 9.0 1.4, 58.9

Trend P = 0.02 Other oral cavity Nonsmokers 6 47 I .o -

1-24 3 23 I .o 0.4, 2.8 25-49 9 31 2.3 0.8, 6.9 so+ 7 26 2. I 0.7,6.8

Trend P = 0.15

206 CANCER January I 1988 Vol. 6 I

TABLE 5. Risk Estimates for Former Cigarette Smokers by Duration of Cessation

Duration of cessation No. of No. of Odds 95% confidence

(Yr) cases controls ratio* limits

Men <5 5-14 15+

Women <5 5-14 I S +

31 17 6.1 2.7. 13.8 10 15 2.2 0.8.6.5 8 26 I .o 0.4, 2.5

Trend P = <0.01

13 4 9.8 2.8, 34.3 3 2 4.5 0.7. 28.0 3 6 1.5 0.3. 7.2

Trend P = <0.01

* Referent category was nonsmokers.

Sixty eight cases (49 men, 19 women) and 70 controls ( 5 8 men, 12 women) had ceased cigarette smoking. Risk estimates were computed for these former smokers by duration of reported cessation (Table 5). For men former smokers whose abstinence was less than 5 years, the risk remained significantly elevated at 6.1. For those who had quit five to 14 years previously, the risk estimate was 2.2, although not statistically significant. For men former smokers who had abstained for I5 or more

TABLE 6. Risk Estimates for Alcohol Consumption (All Study Sites)

No. of No. of Odds 95'70 confidence cases controls ratio* limits

Men Hard liquor

Nonusers Moderate Heavy

Beer/wine Nonusers Moderate Heavy

Women Hard liquor

Nonusers Moderate Heavy

Beer/wine Nonusers Moderate Heavy

16 20 47

22 35 53

5 7

15

12 10 14

28 21 38

Overall

42 31 42

Overall

7 8

15 Overall

20 13 10

Overall

- 1 .o 1.6 0.7, 3.6 2.2 I .O. 4.6 2.0 I .O, 4.2

Trend P = 0.05

- I .o I .8 0.9, 3.4 2.0 1.1. 3.8 I .9 1.0. 3.6

Trend P = 0.0 I

- I .o 0.7 0.1, 14.5 0.6 0.1. 4.1 0.5 0.1 ~ 2.3

Trend P = 0.75

- 1 .o I .3 0.4, 4.1 2.0 0.7. 6.0 I .4 0.5, 3.1

Trend P = 0.19

* Adjusted for smoking status. Moderate: up to twice weekly consumption. Heavy: at least daily consumption.

years previously, the odds ratio was not elevated ( 1 .O). The corresponding risks for women former smokers were 9.8, 4.5, and 1.5, respectively. (These latter risks are based on small numbers and must be treated with caution.) In both instances, the trend for decreasing risk with increased duration of abstinence was statistically significant.

Smoking-adjusted risk estimates for alcohol (a combined beer and wine variable, and hard liquor) for all sites combined are presented in Table 6. Men who de- scribed themselves as moderate drinkers (up to twice weekly consumption) had elevated risks (1.6 for hard liquor, 1.8 for beer/wine). For heavy drinkers (at least daily consumption) these risks were two-fold elevated, and were statistically significant. In both instances, trend analysis also was statistically significant (P < 0.05 and 0.0 I , respectively). For women, a small nonsignificant trend for increasing risk with increasing beer/wine consumption was evident. However, frequency of hard liquor use among women was not reflected in differing risks.

Using nondrinkers who were also nonsmokers as the referent category, risk estimates for various combinations of drinking and cigarette smoking were computed (Table 7). The data were relatively sparse in those categories representing a combination of alcohol use, but no smoking. The measures of effect are therefore somewhat unstable. There were no increased risks associated with alcohol consumption in nonsmoking males and females. Predictably, risk estimates for cigarette smoking in nondrinkers were consistently elevated (two-fold for men, five-fold to six-fold in women). Among men, the combined effect of smoking and alcohol consumption yielded risk estimates above 5 , suggesting a synergistic effect was operative. This interaction was not evident among women.

Multiple logistic regression did not show an independent effect of alcohol, but the interaction term for males suggested that the combined effects of alcohol and tobacco were more than additive, but less than multiplica- tive. There was no such an effect evident among women. However, cell sizes were too small for reliable analysis.

Both cigar use (odds ratio [OR] = 2.8, CL = 1.5, 5.5) and pipe use (OR = 1.8, CL = 1 .O, 3.4) among men were associated with elevated risk for all cancer sites. There was also a significantly elevated risk associated with snuff dipping (OR = 3.4, C1 = 1.0, 10.9). Of the nine cases who dipped snuff, all drank alcohol to some ex- tent, seven also chewed tobacco, and only one denied smoking cigarettes although he had smoked both cigars and pipe at different times. Three of the four controls who dipped snuff also smoked cigarettes and only one denied use of any other tobacco products. There was no difference in distribution of sites of malignancy for snuff

No. 1 RISK FACTORS FOR UPPER AERODIGESTIVE CANCERS - Spitz et al. 207

dippers compared with all other cases. The same number of cases and controls (23) admitted to chewing tobacco so that no significant differences were detectable.

Discussion

The usual methodologic concerns regarding this type of analytic design, derived as it is from a single hospital population, are applicable to our study. Concerns about selection bias were addressed by comparing demo- graphic data for a randomly selected sample of 75 survey respondents and 75 nonrespondents. No differences were detected in the distribution between respondents and nonrespondents of ethnicity, marital status, age, res- idence, or indigent status.

Nonrepresentativeness of the study population arising from differential referral patterns to a tertiary care institution is also of concern. Furthermore, use of a diseased comparison group who might share common exposures with the cases, could bias the odds ratio estimates to- wards unity. Fifteen control patients had bladder cancer, and because of the causal association with tobacco, their inclusion might have reduced the tobacco risk estimates. There is also an overrepresentation of breast cancer and melanoma in the female and male control group, respectively. Breast cancer in particular, with its reputed association with high socioeconomic status, is in direct contrast with head and neck cancer which has an inverse association with social stratification. This could have produced an exaggerated disparity of alcohol and tobacco use between female cases and controls with a re- sultant overestimation of female risks for both tobacco and alcohol use. However, the male cases also were char- acterized by a lower socioeconomic status compared with their controls and thus the constitution of the control group cannot entirely explain the male-female risk differentials.

These limitations are also counterbalanced by the choice of a rigorously defined case group, by the random selection of controls with a diversity of diagnoses, and by the use of a self-administered questionnaire, thereby avoiding biased probing. Besides the obvious logistic advantages, recall bias also is minimized by the choice of a control group composed largely of cancer patient^.^

The etiologic importance of cigarette smoking in risk of oral and laryngeal cancer is without question, and dose-response effects have consistently been demonstrated. Few studies, however, have addressed risk differentials in men and women subjects. Wynder and Stellman documented higher risks for women than men at various smoking/alcohol combinations for laryngeal cancer.' Our data demonstrated higher risks for women smokers at each successive pack-year stratum for all study sites combined. In the heaviest smoking category,

TABLE 7. Risk Estimates for Cigarette Smoking and Alcohol Categories'

Smoking/ No. of No. of Odds alcohol cases controls ratio*

95% confidence limits

Hard liquor

W Y e s Yes/no Yes/yes

No/yes Yes/no Yes/yes

Men

Women

Beer/wine Men

No/yes Yes/no Yes/ yes

No/yes Yes/no Yes/ yes

Women

7 21 60

0 12 22

7 15 81

2 14 22

19 29 40

8 5

15

19 22 54

10 8

13

I .3 2.5 5.3

0.2 6.8 4.2

1.3 2.4 5.3

0.6 5.6 5.4

~~

0.4.4.5 0.9, 1.4 2.0, 14.2

0.1, 3.2 1.7, 21.5 1.3, 13.0

0.4,4.5 0.8,1.3 2.0, 13.9

0.1.4.0 1.5, 21.1 1.6, 18.3

* Referent category was patients who neither smoked nor drank.

the female risk estimate was 12, compared with 7.5 for males.

Not many studies have specifically quantitated tobacco-related risks for oral cavity subsites. Seventy-five percent of oral cavity cancers arise in a horseshoe- shaped area extending from the anterior floor of the mouth and including the lateral lingual margins and tonsillar pillar-retromolar tongue complex.6 It is postu- lated that concentrated carcinogens suspended in saliva are pooled in these mucous reservoirs6 Our smoking risk estimates for oral cavity cancer are consistent with previous An association with heavy smoking also was noted in patients with orohypopharyngeal cancers, with risk estimates not dissimilar from our data." This latter study, however, did not analyze cancers of the tongue and floor of the mouth separately.

Our data also accord well with those from the more numerous studies on smoking and laryngeal cancer. Relative risks ranging from 2 to 3 in the lightest smoking categories to 4 and up to 18.5 in heaviest smoking/alcohol categories are documented.' ' - I 3 Even higher estimates of risk were documented by Wynder et al. (ranging from 6 in lightest smokers to 25 in heaviest smokers).I4 Our comparable estimates were 2.7 and 15.

Wynder and Stellman contended that 10 to 15 years of smoking abstinence were required in long-term men and women smokers before their risks for laryngeal cancer approximated that of nonsmokers,' whereas, Burch ef al. maintained that risk for laryngeal cancer decreased only after 30 years of abstinence." Our data document an 1.5 elevated risk in women who had abstained from cigarette smoking for 15 or more years,

208 CANCER January I 1988 Vol. 61

whereas there was no increase in risk for men who had abstained for the same length of time.

Interaction between alcohol and tobacco in elevating risk of oral, pharyngeal and laryngeal cancer has been repeatedly d e m ~ n s t r a t e d . ~ . ~ ~ . " ~ ' ~ Such an effect is of both biologic and public health significance. Biologi- cally, alcohol could exert a direct local effect, through its solvent properties or by its toxicity to epithelial respiratory enzyme systems." Either mechanism could facili- tate the action of tobacco carcinogens on the target cell. Alcohol also is known to induce microsomal liver en- zymes capable of transforming proximate carcinogens into ultimate carcinogens.'*

However, the risks associated with alcohol consumption, independent of tobacco use, are less clear. Studies of laryngeal cancer reported in 1956,14 1981,'' and 1 98615 found moderately increased risks after adjusting for cigarette smoking. In a 1980 review of reported studies on laryngeal cancer, Rothman ef al." concluded that the data indicated only a modest increase in risk for moderate drinkers, but a strong increase for heavy drinkers. A subsequent investigation of risks for oral squamous carcinoma reported higher relative risks for beer and wine drinkers relative to whiskey drinkers and greater risk for heavy drinkers than for heavy smokers.'' Another case-control study of patients with cancers of the oral cavity, pharynx and larynx also concluded that alcohol had an effect which was independent of, and stronger than, smoking." Our data, in conformity with another case-control study,20 suggest that alcohol exposure, in the absence of cigarette smoking, is not an independent risk factor.

The number of case-control epidemiologic investiga- tions evaluating the relationship between smokeless tobacco and cancer risk is not large, and many have methodologic limitations.2' However, patterns of increased risk for oral cancer have been fairly consistent across studies (both national and international) with risks ranging from 1 .4-fold to 48-fold for gum lesions in heaviest snuff Unlike the oral cavity, the larynx comes into contact with constituents of smokeless tobacco in more dilute concentrations. Results of studies on laryngeal cancer in relation to smokeless tobacco have been i n c o n c l ~ s i v e ~ ~ . ~ ~ (four of nine dippers in our series had laryngeal cancer, four were oral cavity cancers and one had an orohypopharyngeal lesion), a distribution not dissimilar from the overall case group.

These data emphasize the importance of combining strategies for alcohol and tobacco cessation, particularly in women smokers, whose risks are higher and whose success rate in achieving and maintaining abstinence is lower.2' Finally, the differences noted by gender and by primary site in both the magnitude of the risk estimates and in the gradients of the dose-responses, suggest a

variable susceptibility to carcinogenic action, which may be sex and site dependent.

REFERENCES 1. Epilog Software-Version 3-0. Epidemiology and clinical trials

statistical package. Epicenter Software, P. 0. Box 90073, Pasadena, California 9 I 109.

2. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies ofdisease. J Nail Cancer Inst 1959; 22:719- 748.

3. McGee DL. A program for logistic regression on the IBM PC. Am J Epidemiol 1986; 124702-705.

4. Linet MS. Brookmeyer R. Use ofcancer controls in case-control cancer studies. Am J Epidemiol 1987; 125:l-1 I.

5 . Wynder EL, Stellman SD. Impact of long-term filter cigarette usage on lung and larynx cancer risk: A case-control study. J Nail Cancer Insr 1979; 62:471-477.

6. Moore C. Catlin D. Anatomic origins and locations of oral cancer. Am JSurg 1967; I14:510-513.

7. Graham S, Dayal H, Rohrer T el al. Dentition, diet, tobacco, and alcohol in the epidemiology of oral cancer. J Null Cunccr Inst 1977;

8. Rothman K. Keller A. The effect ofjoint exposure to alcohol and tobacco on risk of cancer ofthe mouth and pharynx. JChron Dis 1972;

9. Jayant K, Balakrishnan V. Sanghvi LD er al. Quantification of the role of smoking and chewing tobacco in oral, pharyngeal. and esophageal cancers. Br J Cancer 1977; 35:232-235.

10. Elwood JM, Pearson JCG, Skippen DH. Jackson SM. Alcohol, smoking, social and occupational factors in the aetiology of cancer of the oral cavity, pharynx and larynx. Int J Cancer 1984; 34:603-612.

I I . Burch JD, Howe GR. Miller AB, Semenciw R. Tobacco, alcohol, asbestos, and nickel in the etiology of cancer of the larynx: A case-control study. J NailCancer Inst 1981; 67:1219-1224.

12. Flanders WD, Rothman KJ. Interaction ofalcohol and tobacco in laryngeal cancer. Am J Epidemiol 1982; I15:371-379.

13. Stell PM. Smoking and laryngeal cancer. Lancer 1972; 1:617- 618.

14. Wynder EL, Bross IJ, Day E. A study of environmental factors in cancer of the larynx. Cancer 1956; 9236-1 10.

15. Zagraniski RT, Kelsey JL, Walter SD. Occupational risk factors for laryngeal carcinoma: Connecticut, 1975-1980. A m J Epidemiol

16. Rothrnan KJ, Cann CI. Flanders D, Fried MP. Epidemiology of laryngeal cancer. Epidemiol Rev 1980 2: 195-209.

17. McCoy GD, Hecht SS, Wynder EL. The roles of tobacco, alcohol, and diet in the etiology of upper alimentary and respiratory tract cancers. P r o Med 1980; 9:622-629.

18. Vitale JJ, Broitman SA, Gottlieb LS. Alcohol and carcinogen- esis. In: Newell GR, Ellison NM. eds. Nutrition and Cancer: Etiology and Treatment. New York: Raven Press, 1981; 29 1-302.

19. Mashberg A, Garfinkel L. Hams S. Alcohol as a primary risk factor in oral squamous carcinoma. CA 1981; 31:146-155.

20. Winn DM, Ziegler RG, Pickle LW, Gridley G, Blot WJ, Hoover RN. Diet in the etiology of oral and pharyngeal cancer among women from the southern United States. Cancer Res 1984; 441216-1222.

2 I . The health consequences of using smokeless tobacco: A report ofthe Advisory Committee to the Surgeon General. US Dept of Health and Human Services, NIH Pub1 #86-2874, Bethesda, MD, 1986.

22. Winn DM, Blot WJ, Shy CM, ei al. Snuff dipping and oral cancer among women in the southern United States. N Engl J Med

23. Williams RR, Horm JW. Association of cancer sites with tobacco and alcohol consumption and socioeconomic status of patients: Interview study from the Third National Cancer Survey. J Natf Cancer Inst 1977; 58525-547.

24. Wynder EL, Stellman SD. Comparative epidemiology of tobacco-related cancers. Cancer Res 1977: 37:4608-4622.

25. Jackson PH, Stapleton JA, Russell MAH, Memman RJ. Pre- dictors of outcome in a general practitioner intervention against smoking. Prev Med 15; 1986:244-253.

591161 1-1618.

25:71 1-716.

1986; I24:67-76.

1981; 304:745-749.

Documents

Squamous cell carcinoma of the upper aerodigestive tract. A case comparison analysis