Nasopharyngeal Cancer in a Low-Risk Population: Defining ...Nasopharyngeal cancer (NPC) is a major public health problem in parts of Southeast Asia and North Africa, but is rare among

Vol. 5. 587-593. August 1996 Cancer Epidemiology, Biomarkers & Prevention 587

Nasopharyngeal Cancer in a Low-Risk Population: Defining Risk

Factors by Histological Type’

Thomas L. Vaughan,2 Jean A. Shapiro, Richard D. Burt,G. Marie Swanson, Marianne Berwick, Charles F. Lynch,and Joseph L. Lyon

Prograni in Epidemiology. Fred Hutchinson Cancer Research Center. Seattle,Washington 98104 T. L. V.. J. A. S.. R. D. B.I: Department of Epidemiology.

University of Washington. Seattle. Washington 98195 IT. L. V., J. A. S..

R. D. B.J: Cancer Center and College of Human Medicine. Michigan State

University. East Lansing. Michigan 48824 10. M. 5.1: Department of

Epidemiology and Public Health, Yale University. New Haven, Connecticut(16510 IM. B.I: Memorial Sloan-Kettering Cancer Research Center. New York.

New York IO()21 lM. B.l: Department of Preventive Medicine and

Environmental Health. University of Iowa. Iowa City. Iowa 52242 IC. F. Li;

and Department of Family Medicine. University of Utah. Salt Lake City. Utah

84132 J. L. L.I

Abstract

Nasopharyngeal cancer (NPC) is a major public healthproblem in parts of Southeast Asia and North Africa, butis rare among whites and blacks. Although infection withthe EBV and genetic susceptibility appear to play largeroles in high-incidence populations, migrant studiessuggest that environmental factors may also beimportant. Aside from the high risks associated withingestion of salted fish, surprisingly few other risk factorshave been established from studies in endemic areas. Westudied a low-incidence population to determine whethertobacco use, alcohol consumption, and certain medicalconditions and treatments are related to NPC and toexamine variations in risk by histology. We reasoned thatnew relationships might be best identified in the absenceof strong causal pathways, such as intake of preservedfoods and genetic susceptibility. A population-based case-control study was conducted from 1987 to 1993 at fivecancer registries in the United States: westernWashington, metropolitan Detroit, Connecticut, Iowa, andUtah. Controls were identified by random digit dialingand frequency matched to the gender and agedistribution of cases at each registry. Telephoneinterviews were completed by 231 cases and 246 controls.We observed a strong dose-response relationship betweencigarette smoking and risk of differentiated squamous cellcarcinoma (test for trend, P < .001). The highest risk[odds ratio (OR), 6.5; 95% confidence interval (CI), 2.0-21.3] occurred among current smokers with a history of

Received 2/I 5/96: revised 4/24/96: accepted 4/30/96.The costs of publication of this article were defrayed in part by the payment of

page charges. This article must therefore be hereby marked adt’ertise,nent in

accordance with 1 8 U.S.C. Section 1734 solely to indicate this fact.

� This research was supported in part by NCI grants R29 CA46552 and R03

CA50256. and NCI contract NOl CN05230.2 To whom requests br reprints should be addressed. at Program in Epidemiology

(MP474). Fred Hutchinson Cancer Research Center. I 124 Columbia Street

Seattle. WA 951(14. Telephone: 206-667-4741: FAX: 206-667-4787.

more than 60 pack-years. In contrast, there was noevidence that undifferentiated or nonkeratinizingcarcinomas were associated with cigarette smoking.Similarly, a significant increase in risk was observed forthe heaviest alcohol consumers (21 or more drinks/week)only for differentiated squamous cell carcinomas (OR,2.9; 95% CI, 1.2-6.9). The associations with cigarettesand alcohol appeared to be stronger among persons 50years or older. There was a suggestion that diagnosiswith infectious mononucleosis (a marker of late infectionwith EBV) is linked with decreased NPC risk (OR, 0.4;95% CI, 0.1-1.1). This report indicates that over two-thirds of differentiated squamous cell NPC cases arisingin older persons in the United States can be accountedfor by cigarettes and alcohol, but leaves unexplainedcases arising in the young and carcinomas ofundifferentiated or nonkeratinizing histology. Futurestudies of NPC need to take into account histology andage in evaluating these and other environmental andgenetic risk factors.

Introduction

The incidence of NPC3 varies considerably throughout theworld. In Southern China, other parts of Southeast Asia, North

Africa, and among Eskimos, NPC is of major public health

importance, with annual incidence rates as high as 20/1()0,(XX)

(I). In contrast, NPC is a much rarer cancer among whites

(0.4/100,000) and blacks (0.6/100,000) in the United States and

elsewhere (2). The histological distribution varies along with

incidence rates, with high-risk populations having a larger

proportion of more poorly differentiated carcinomas (2).

A link between nasopharyngeal carcinoma and infectionwith the EBV is well documented (3, 4). However, infection

with EBV is a worldwide phenomenon; serological evidence

points to its almost ubiquitous presence. Thus, even though the

EBV infection appears to be a necessary factor in the develop-ment of many nasopharyngeal carcinomas, it is also clear that

additional cofactors are required before a malignancy is

expressed.Genetic susceptibility is likely to be one such cofactor; for

example, high-risk HLA types have been identified in bothhigh- and low-incidence populations, and familial clustering

has been observed (5-7). However, there is also a growing bodyof evidence that environmental exposures can cause NPC. The

decreasing rate of cancer among successive generations of

Chinese migrants to Hawaii and California and the changes in

risk among migrants between high- and low-risk areas of

China, along with the increased risk of NPC among people

, The abbreviations used are: NPC. nasopharyngeal cancer: SEER. Surveillance.

Epidemiology. and End Results: ICD-O. International Classification of Disease-

Oncology: OR. odds ratio: CI. confidence interval: NOS. not otherwise specified.

on May 14, 2021. © 1996 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

http://cebp.aacrjournals.org/

588 NPC in a Low-Risk Population

otherwise at low risk who are born and raised in high-risk areas,

suggest that external factors. some of them operating early inlife, are important risk determinants (8-13).

One of the strongest environmental risk factors identifiedthus far is eating salted fish, especially during childhood (14).

Other studies have inconsistently linked increased NPC riskwith cigarette smoking, alcohol intake. low intake of fruits andvegetables, and certain occupations ( I ). However, most epide-

miological studies of environmental factors have focused on

populations at high risk of the disease. Studying a populationwhich is inherently at low risk may facilitate detecting addi-tional risk factors, which in a high-risk population would oth-

erwise be overwhelmed by the action of other causal pathways

such as childhood diet and genetic susceptibility. In 1988, webegan a collaborative population-based case-control study ofNPC at five cancer registries in the United States. In this secondreport from the project (7). we examine the relative risk asso-

ciated with tobacco exposure, alcohol consumption, and certainmedical conditions.

Subjects and Methods

Cases were prospectively identified by five population-based

cancer registries participating in the National Cancer Institute’s

SEER Program. Eligible persons included males and females ofany race between I 8 and 74 years of age who were diagnosed

with any histological type of NPC (ICD-O code 147). At fourregistries (Connecticut. metropolitan Detroit, Iowa, and Utah),cases diagnosed between April 1 , 1987 and June 30, 1991 wereincluded; at the western Washington registry, case ascertain-

ment was extended to June 30. 1993.

Controls were identified by random digit dialing, using amodification of Waksberg’s method (15). and frequencymatched to the gender and age (in 5-year groups) distribution ofthe cases at each registry. The protocol was as follows. From alist of all working exchanges (based on information provided

from the local telephone companies), an exchange was ran-domly selected and four random digits appended, forming a

telephone number. This number was then called up to nine

times, at different times of the day and week, to determinewhether the number was for a residence. For numbers selected

in this manner that identified a residence, the first five digits(plus area code) were used to generate additional phone num-bers by appending two random digits. Secondary numbers

generated in this way were then resolved one by one until twoadditional residences were reached.

Trained interviewers at each registry conducted structuredtelephone interviews with cases and controls that inquired about

demographics, the history of certain medical conditions andtreatments, history of exposure to tobacco and alcohol products,and occupational and residential histories. A standardized self-administered questionnaire was used to collect information on

dietary habits. All questions referred to the time period before

the reference date, which was 1 year before diagnosis for cases

and I year before ascertainment for controls.

We identified 294 eligible cases. The distribution by reg-istry was: western Washington. 105 cases; Detroit, 75 cases;

Connecticut, 64 cases; Iowa, 36 cases; and Utah, 14 cases. Ofthese subjects, interviews were successfully completed for 240(82C/�), The two most common reasons for noninterviews were

subject and physician refusal. Demographic information oncases who were not interviewed was gathered from medical

records as part of the cancer registry operations. To maintain

comparability with controls, nine cases without a telephone atthe reference date were excluded from further analyses, leaving

23 1 cases to be analyzed. In ascertaining eligible controls, 2649

(92%) of 2885 phone numbers were successfully screened. Thisprocess identified 324 eligible subjects, of whom 246 (76%)

completed interviews, yielding a 70% (0.92 X 0.76) overallresponse rate. Two controls who did not have a telephone at thereference date were excluded, leaving 244 to be analyzed.Interviews with the closest next-of-kin were completed in in-

stances where a subject was too ill or deceased; this occurred

for 44 ( I 9%) of the interviewed cases and 3 ( I %) of the

controls.Logistic regression was used to calculate ORs and 95%

CIs ( I 6). The matching variables, SEER site, age (categorizedinto nine groups), and gender were taken into account in all

analyses by using them to define matched sets for conditionalanalyses. Trend tests were calculated using the likelihood ratio

statistic associated with adding the risk factor of interest in

grouped linear form (i.e. , consecutive integers) to a model

containing the other risk factors. Tests for homogeneity werecalculated using the likelihood ratio statistic associated with the

addition of an interaction term to the model. In some instances,stratified analysis was also used to calculate a conditional

maximum likelihood estimate of the OR and an exact 95% CI(17). Population-attributable risk percentages were calculatedusing adjusted ORs from the logistic regression analyses (18).

Results

Selected demographic and clinical characteristics of the inter-viewed cases, noninterviewed cases, and controls are given inTable I . Histological types were grouped on the basis of ICD-Ocodes to be similar to the WHO classification scheme (see

Table 1 , footnote a; Ref. 19). Approximately one half of allcases were classified as differentiated squamous cell carcino-mas, with undifferentiated carcinomas, including anaplastic and

lymphoepithelial tumors, being the next most common. Thenoninterviewed cases were similar in histology and age to thosewho were interviewed, but somewhat more likely to be femaleand of the black or Asian race. They were also more likely tohave distant stages of disease at diagnosis. As designed, the

interviewed cases were quite similar to the controls with regardto age and gender; however, persons of the Asian race and thoseborn outside of the United States were much more frequent

among the cases. The mean age of interviewed cases by racewas 56.8 years for whites, 48.9 years for blacks, 42.6 years for

Asians, and 43.5 years for American Indians and Eskimos.

The histology-specific ORs associated with cigarettesmoking, alcohol use, and education are given in Table 2. A

clear dose-response relationship between cigarette smoking andrisk of differentiated squamous cell carcinoma of the nasophar-ynx is evident. For current smokers with a history of 60 or morepack-years, the OR 6.5 (95% CI, 2.0-21.3). A similar

association with patterns of cigarette use was found for cancers

classified as epithelial, NOS. In contrast there was no evidencethat undifferentiated or nonkeratinizing carcinomas (analyzedseparately or together) were associated with cigarette smoking.Lymphomas were also unrelated to cigarette smoking (data notshown). Direct comparison of differentiated cancers to undif-

ferentiated and nonkeratinizing cancers revealed that the asso-ciations with cigarette smoking were significantly different

(P = 0.002) for the two histological groups, whereas the

associations with alcohol were compatible with a common OR

(P = 0.15).

For differentiated squamous cell carcinomas, drinking 21

or more alcoholic drinks per week was associated with a 2.9-fold (95% CI, 1.2-6.9) increase in risk, whereas no increase in



Table / Selected characteristics of cases and controls

Controls Interviewed Noninterviewedcases cases

No. No. No.

(n�244) (n=231) (n54)

Histology”

Epithelial. NOS

Undifferentiated

Nonkeratinizing

Differentiated squamous

Lymphoma

Other/unknown

Stage

Local

Regional

Distant

Unknown

Age (yr)

Mean (SD)

18-39

40-49

50-59

6t)-69

70-74

Gender

Male

Female

Race

White

Black

Asian

American Indian/Eskimo

Unknown

Country of birth

United States

Pacific Islands

Asia

Other

55.2 (13.0) 54.2 (13.4) 54.4 (14.9)

32 13.1 34 14.7

42 17.2 46 19.9

62 25.4 50 21.6

76 31.1 73 31.6

32 13.1 28 12.1

164 67.2 154 66.7 32 59.3

80 32.8 77 33.3 22 40.7

224 91.8 195 84.4

I 0.4 6 2.6

0 0.0 13 5.6

19 7.8 17 7.4

“ ICD-O classifications: epithelial. NOS (801x-804x, except 8020 and 8021);

undifferentiated (8020. 8021. and 8082): nonkeratinizing (8072 and 8073); dif-

ferentiated squamous (805x-808x. except 8072, 8073, and 8082); lymphoma(959x-969x); other and unknown (all others. including 800x).

risk was observed for less frequent intake. Again, we found asimilar association with epithelial, NOS cancers, and no asso-

ciation with undifferentiated or nonkeratinizing carcinomas orwith lymphomas. For all histological types examined, there wasa strong inverse relationship between increasing years of edu-

cation and NPC risk.When proxy interviews were excluded in analyses of dif-

ferentiated squamous cell carcinomas, the associations with

cigarettes and alcohol were slightly attenuated. The OR for

current smokers decreased from 6.5 to 5.8, and the OR for 21or more drinks per week decreased from 2.9 to 2.0. Adjustingfor usual amounts of fruit and/or vegetable intake did notmeaningfully change the risk estimates for cigarettes, alcohol,and education. There was no statistical evidence that the joint

effects of cigarette smoking and alcohol intake were differentthan would be expected from a multiplicative model (test forhomogeneity, P = 0.53), although the OR for drinking 21 ormore drinks per week was somewhat higher among nonsmokers

and former smokers (OR, 3.8; 95% CI, 1.4-10.3) than amongcurrent smokers (OR, 2.3; 95% CI, 0.7-7.8). There was littledifference in the associations with cigarettes and alcohol by

Cancer Epidemiology, Biomarkers & Prevention 589

gender.We observed little difference in the associations with

cigarette smoking and alcohol intake when the Asian subjectswere excluded from analysis. Unfortunately, the scarcity of

Asian controls prevented more detailed analysis of race-spe-

cific associations. Controlling for race among non-Asian sub-jects did not change the relative risk estimates. The relative riskestimates for cigarette and alcohol use among blacks were

similar to those for whites, but the CIs were quite wide.24 10.4 2 3.7 We also observed that the association of differentiated37 16.0 10 18.5 squamous cell carcinomas with cigarettes appeared to be stron-17 7.4 3 5.6 ger among persons 50 years or older than it was for younger

I 18 51.1 28 51.9 persons. The ORs for ever smoking were 4.0 (95% CI, 1.7-9.3)19 8.2 7 13.0

16 6 9 4 7 4 and 1 . 1 (95% CI, 0.3-4.3) for the older and younger age. groups, respectively (test for homogeneity, P = 0. 16). The

45 19.5 8 14.8 pattern by age was similar for alcohol but with wider CIs; theI 16 50.2 25 46.3 ORs were 3.3 (95% CI, 1 .4-7.8) and 1 .4 (95% CI, 0. 1-15.1),

52 22.5 19 35.2 respectively.18 7.8 2 3.7 Population-attributable risk percentage estimates are given

in Table 3. Cigarettes and alcohol together accounted for 38%of all epithelial carcinomas, 61% of differentiated squamous

9 16.7 cell carcinomas, and 70% of differentiated squamous cell car-

15 278 cinomas ansing in persons 50 years old or older.13 241 To examine the associations between risk of differen-

8 14.8 tiated squamous cell carcinomas and cessation of cigarettesmoking and type of cigarette smoked, we carried out addi-

tional analyses after excluding persons who had never reg-ularly smoked cigarettes (Table 4). There was a substantial

and rapid decrease in risk with cessation. When the type of221 90.6 178 77.1 33 61.1 cigarette was examined, we found little difference in risk

21 8.6 24 10.4 7 13.0 between persons who were lifetime nonfilter cigarette smok-

: � 25 10.8 10 18.5 ers and those who began smoking nonfilter cigarettes andt) 00 0 00 3 56 switched to filter cigarettes. Both groups experienced ap-. proximately twice the risk as persons who were lifetime

filter cigarette users after controlling for the frequency and

duration of cigarette smoking. There was no risk differential

between persons who usually smoked filter cigarettes versus

those who usually smoked nonfilter cigarettes.In analyses restricted to persons who had ever regularly

drunk alcoholic beverages, there was no significant differencein differentiated squamous cell NPC risk by usual type of

alcoholic beverage after controlling for the usual amount drunk(Table 4).

The association between differentiated squamous cellNPC and residential exposure to environmental tobacco smokewas assessed among lifetime nonsmokers and also among for-mer smokers (data not shown). We found no evidence that the

presence of one or more smokers in the household duringchildhood or adulthood increased risk. The results were similarwhen the number of other smokers in the household or the

smoker-years (sum of the number of years each member of the

household smoked) were considered.There was no association between the use of cigars or

pipes and differentiated squamous cell NPC risk when all malesubjects were included in the analyses. (There were no femalepipe or cigar smokers.) When current cigarette smokers were

excluded, the ORs for regular cigar and pipe use (for 6 monthsor longer) were elevated (OR, 3.7; 95% CI,= 0.9-14.9 and OR,

2.1; 95% CI, 0.6-7.6, respectively). When subjects under theage of 50 were also excluded, the OR for cigar use decreased

to 2.6 (95% CI, 0.5-12.3), and the OR for pipe use increased to2.7 (95% CI, 0.7-10.4).

The ORs associated with several medical conditions andtreatments potentially related to risk of NPC are given in Table

5. There was a suggestion that diagnosis with infectious mon-onucleosis is linked with decreased NPC risk (OR, 0.4; 95% CI,

0.1-1.1). The mean age of diagnosis of infectious mononucle-



590 NPC in a Low.Risk Population

Table 2 OR” for NPC associated with cigarette smoking. alcohol use, and education

Controls Cases

All Undifferentiated Differentiated .. . . . . Epithelial NOS

No. epithelial and nonkeratinizing squamous cell

(n 244) No No No No

(it = 196) OR 95% CI (,t = 54) OR 95% CI (,t = 118) OR 95% CI (a 24) OR 95% CI

Cigarette use

Never 95 47 1.0 22 1 .0 19 1 .0 6 1.0

Former 98 61 1.3 0.7-2.3 20 0.9 0.4-2.1 32 1.7 0.8-3.6 9 2.5 0.7-8.6

Current: 1-4 pack-yr 24 31 1.9 0.9-4.0 7 0.5 0.2-1.9 19 5.7 2.1-15.8 5 1.4 0.3-6.7

Current: 35-59 pack-yr 20 27 3.0 1.3-6.8 3 0.6 0.1-3.1 22 5.1 2.0-13.3 2 3.0 0.4-21.5

Current: 60+ pack-yr 7 23 4.3 1.5-12.4 1 0.5 0.0-5.7 20 6.5 2.0-21.3 2 5.5 0.5-58.4

Unknown 0 7 1 6 t)

Trend test” P < 0.(X)I P = 0.30 P < 0.001 P = 0.19

Alcohol use (drinks/week)

0-6 162 105 1.0 35 1.0 70 1.0 14 1.0

7-13 35 25 1.0 0.6-1.9 8 1.5 0.5-4.0 17 0.9 0.4-2.0 3 1.0 0.2-4.6

14-20 19 14 0.9 0.4-2.0 4 1.0 0.3-3.8 10 0.9 0.3-2.8 3 1.6 0.3-7.5

21+ 21 31 2.0 1.0-4.0 5 1.3 0.3-5.1 26 2.9 1.2-6.9 4 3.0 0.7-8.6

Unknown 7 21 2 19 0

Trend test” P = 0. 15 P = 0.67 P = 0.04 P = 0.16

Education completed

Less than high school 34 55 1.0 10 1.0 45 1.0 5 1.0

High school 73 7t) 0.4 0.2-418 19 0.3 0.1-1.0 51 0.5 0.2-1.0 10 0.5 0.1-2.2

More than high school 137 70 0.3 0.1-0.5 25 0.2 0.1-0.6 45 0.3 0.1-0.6 9 0.2 0.0-0.9

Unknown 0 I 0 1 0

Trend test” P < 0.001 P = 0.004 P - 0.005 P = 0.t)2

“ ORs adjusted for age. gender. SEER site, and other risk factors as categorized.

S Trend test based on likelihood ratio statistic associated with adding risk factor (as categorized) in grouped linear form to logistic model.

Table 3 ORs” and population-attributable risks for NPC by histology and age

All epithelial Differentiated squamous cell

All ages All ages 50+ Years- -

No. OR 95’7 CI PAR (%)b No. OR 95% CI PAR (%) No. OR 95% CI PAR (%)

control/case control/case control/case

Cigarette use 34 55 65

Never 95/47 1.0 95/19 1.0 62/13 It)

Former 98/61 1.3 0.7-2.2 98/32 1.6 0.8-3.5 75/30 2.5 1.0-6.2

Current 51/81 2.6 1.4-4.6 51/61 5.6 2.6-12.0 33/45 7.6 2.9-20.0

Alcohol intake 9 15 17

<21 per week 216/144 1.0 216/77 1.0 146/59 1.0

21+ perweek 21/31 2.1 l.()-4.1 21/22 3.1 1.3-6.9 17/20 3.2 1.3-7.8

Cigarettes and alcohol 38 61 70

“ ORs adjusted for education. age. gender. SEER site. cigarette use (three categories), and alcohol use (two categories).

“ PAR. population-attributable risks.

osis among the 14 controls was 18.9 years (SD, 8.1) compared

to 25.4 years (SD, I 2.6) among the 5 epithelial NPC cases (testfor difference in means, P = 0.20). These cases included two

undifferentiated tumors, one nonkeratinizing tumor, and twodifferentiated squamous cell tumors. The ORs associated witha diagnosis of mononucleosis were quite similar when subjectswere stratified by age and when Asians were excluded from

analyses.We observed no association between epithelial nasopha-

ryngeal carcinoma and a diagnosis before the reference date of

nasal polyps, sinusitis, or repeated nose bleeds (Table 5). Fur-

thermore, there was no indication of increased risk with ahistory of childhood radiation treatment to the head and neckfor reasons such as an enlarged thymus and acne or with regular

use of nasal sprays or nose drops.Among lymphoma cases, 4 of 19 persons reported regular

use of nasal sprays or nose drops. After controlling for educa-tion (three levels) and age (<50 versus � 50) in a stratifiedanalysis, the OR was 3.5 (exact 95% CI. 0.7-13.2). However,there was not a dose-response relationship: the OR associatedwith heavy cumulative use (>2000 times) was 2.4 compared to5.2 for <2000 times.

Discussion

We found quite different risk factor profiles for NPC by his-

tological type. In particular, cigarette smoking and heavy alco-hol intake were unrelated to the risk of undifferentiated and

nonkeratinizing squamous cell carcinomas, but were importantrisk factors for differentiated squamous cell cancer, accounting

for approximately three of every five such cases.The current study has several strengths: (a) it is based on




Table 4 ORs for differentiated squamous cell NPC associated with selected

smoking and alcohol variables

Controls Differentiated squamous cell

No. No. OR 95% CI

Recency of cigarette use”

Still smoking 51 66 1.0

Quit <5 yr 19 9 0.1 0.0-0.6

Quit 5-14 yr 30 1 1 0.2 0.1-0.7

Quit 15+ yr 49 12 0.2 0.0-0.8

Unknown 0 0

Trend test” P = 0.003

Type of cigarette ever smoked’

Filter 45 21 1.0

Mixed 71 61 2.3 0.8-6.6

Nonfilter 33 13 2.0 0.5-7.8

Unknown 0 3

Usual type of cigarette’

Filter 89 59 1.0

Nonfilter 57 30 0.9 0.4-2.2

Unknown 3 9

Usual type of alcohol”

Beer 75 31 1.0

Wine 31 10 1.6 0.4-6.4

Liquor 55 24 0.9 0.4-2. I

Unknown 5 1

“ Among known ever-smokers, adjusted for alcohol use, frequency of cigarettesmoking. duration of smoking, education, age. gender. and SEER site.

“ Trend test based on likelihood ratio statistic associated with adding risk factor

(as categorized) in grouped linear form to logistic model.

‘ Among known ever-smokers. adjusted for alcohol use, cigarette use, education,

age. gender, and SEER site.

,, Among known ever-drinkers, adjusted for alcohol use, cigarette use. education,

age. gender, and SEER site.

the largest reported series of cases occurring in a low-riskpopulation; (b) it is population-based; (c) it is based on incidentcases rather than deaths; and (d) detailed histological informa-

tion is available. These attributes, combined with the consistentdose-response relationship and decrease in risk associated withsmoking cessation, add strong support for a causal role ofcigarette use in differentiated squamous cell cancer. The patternof elevated risk associated with alcohol consumption is lessstriking, but also suggests a causal relationship. Although it ispossible that residual confounding from the effects of tobacco

may account for a part of the association with alcohol, the

observation that alcohol intake was most strongly associatedwith NPC risk among former and nonsmokers implies that this

is not an important limitation. The lack of association betweencigarette and alcohol use and risk of less differentiated carci-nomas is consistent with the possibility that EBV plays a largerrole in such tumors ( I , 7, 20).

We also observed that cigarette use, and to a lesser extent,

alcohol consumption appeared to be related to squamous cellNPC risk primarily among persons older than 50 years, ac-counting for 70% of such cases. This is in contrast to lung

cancer, for which cigarette use increases risk across all ages(2 1 ). This finding suggests that squamous cell carcinomas of

the nasopharynx occurring among younger people in the United

States may have a very different etiology than those in olderpeople. Although chance may explain this finding, another

possible explanation is that early childhood exposures mayunderlie a much larger proportion of cancers occurring in the

younger age groups (22). In high-incidence areas of China,there is good evidence for an important role of early exposureto certain foods, such as salted fish ( I 4). However, little is

known about childhood exposures and NPC risk in lower in-

cidence areas.

Another possibility is that inherited susceptibility may

account for many of the cancers in the under 50 age group. Ina study from Taiwan (23), 17 of 20 (85%) cases with a familyhistory were less than 50 years old, compared to 172 of 327

(53%) sporadic cases. Very few of the subjects in the currentstudy reported a family history of NPC; therefore this marker ofsusceptibility could not be examined.

The apparent effect modification by age may explain, atleast in part, the conflicting evidence for the role of cigarettes,

and largely negative evidence for alcohol, that has been re-ported in the literature from high- and moderate-incidenceareas, as reviewed by Hildesheim and Levine ( I ). For example,Chen et a!. (23) reported a small overall elevation in risk (OR,1.6; 95% CI, = 0.8-3.4) among smokers of 20 or more ciga-

rettes per day in Taiwan. However, stratification by age re-

vealed ORs at or below one for persons less than 50 years oldand ranging from approximately 3 to 8 among persons older

than 50 years. In contrast, Yu et a!. (24), in a study of 306incident cases of NPC occurring among residents younger than50 years in a high-risk area of Southern China, reported an

almost 3-fold increase in risk among the heaviest smokers.Other studies have not reported whether age modified theassociation with cigarette smoking or alcohol; thus this issueremains unresolved.

Another explanation of the absence of consensus in theliterature regarding smoking and alcohol is the lack of account-

ing for histological type in previous studies. The proportion ofundifferentiated and nonkeratinizing carcinomas is higher inSouthern Asia and other high-incidence areas (2); therefore, theoverall association with alcohol and cigarettes would be atten-

uated in studies of these populations that ignore histological

type. Finally, differences in study design, particularly the use ofhospital-based controls, may also have played a role in some ofthe negative studies of cigarette smoking, alcohol and NPC risk(25, 26).

When studies of U.S. populations are considered, the as-sociation with cigarettes appears to be more consistently pos-itive (25, 27, 28). The strongest previous evidence for a role for

alcohol also comes from a study of U.S. whites (27). Aftercontrolling for cigarette smoking, they observed an OR of 1.8

(95% CI, 1 .1-3.1) among persons consuming 24 or more drinks

per week and no increase in risk for less frequent intake. Studiesin high-incidence areas have generally found no association,

although most did not report detailed analyses which separated

out the heaviest drinkers (I). Possible mechanisms underlyingthe well-accepted association with oral and pharyngeal cancers

have been reviewed elsewhere (29). In addition, animal studieshave suggested that alcohol intake may enhance the carcino-

genicity of certain ingested nitrosamines (30, 31), which in turnmay underlie the carcinogenicity of salted fish and other pre-served foods linked to NPC in high-risk areas (32, 33). How-

ever, in light of the evidence thus far that alcohol is not a riskfactor for NPC in such areas, particularly for the undifferenti-ated carcinomas which are more common there, this does notappear to be a clinically important mechanism.

We also found a suggestion of increased risk associated

with cigar and pipe smoking among persons who were notcurrent cigarette smokers, although this finding could easily

have occurred by chance. Chow et a!. (25) reported no increasein risk associated with cigar or pipe use, based on only two such

exposed cases. After controlling for the intensity, duration, and

current smoking habits of persons who had ever smoked cig-arettes, we found that lifetime filter cigarette users experienced



592 NPC in a Low.Risk Population

Table 5 ORs for epithelial N PC associated ss ith selected me dical variables

Controls

No

(a 244)

All epithelial

No

(it 196)

Demographics adjustment” Full adjustment”

OR’ 95% CIOR’ 95’f Cl

Ever diagnosed or treated by a doctor for the following conditions

Infectious mononucleosis 4 5 0.3 t).l-l.O 0.4 0.1-1.1

Nasal polyps 5 6 1.5 t).4-5.O 1.6 0.4-5.9

Sinus infection or sinusitis 66 52 0.9 0.6-1.5 1.0 0.6-1.6

Repeated nose bleeds 20 15 0.8 t).4-I.6 0.8 0.4-1.8

Childhood radiation treatment to head or neck”

Yes 4 2 0.8 t).l-5.3 0.8 t).l-5.7

Ever used nose drops or nasal sprays on a regular basis for at least I yr

Yes 17 13 0.8 0.4-1.7 0.4 0.2-1.2

Total no. of times nose drops or nasal sprays used

Regular user, <2(XX) times 9 4 0.5 0.2-1.8 0.2 0.0-1.0

Regular user. 20(X)+ times 7 6 0.7 t).2-2.4 0.5 0.1-2.1

“ Adjusted for age. gender. and SEER site.

‘, Adjusted for alcohol use, cigarette use. education. age. gender. and SEER site.‘ Compared to subjects who were never diagnosed with or treated for the condition.

.1 Through the age of I 8 years.

approximately one half the risk of lifetime nonfilter users, andthat there was no evidence of a beneficial effect of switching tofilter cigarettes. No other studies have examined this issueregarding NPC.

When lifetime nonsmokers were analyzed we found no

evidence of increased risk associated exposure to environmen-

tal tobacco smoke at home as a child or adult. Yu ci al. (34)initially reported a 2-fold increase in risk associated with a

household member smoking around the time of birth in a studyof persons less than 45 years old in a high-incidence area ofSouthern China. However, a subsequent study, also amongyounger persons in Southern China, did not confirm this result(24). As noted above, a strength of the present study is its

ability to focus on histological types and age groups moststrongly associated with smoking in evaluating this relation-

ship. However. such focus also reduces the number of subjects

available for analysis; thus the negative finding with residentialexposure to cigarettes must be viewed in light of the low powerto detect an association.

Because of the relatively poor prognosis for NPC, a sub-

stantial proportion of case interviews (19%) were conductedwith next-of-kin. Exclusion of these interviews from analysesresulted in slighter weaker associations with smoking and al-

cohol. This could imply an overestimation of cigarette andalcohol use by proxy respondents. Alternatively it would beconsistent with a shorter survival (and less likelihood of beingdirectly interviewed) of cases who frequently used cigarettesand alcohol. Such a pattern of decreased survival has beennoted for other head and neck cancers (35).

We observed a nonsignificant 60% decrease in epithelialNPC risk associated with a diagnosis of infectious mononucle-osis, a marker of late infection with EBV. This suggests an

increased risk of NPC with earlier age at infection. This obser-vation is based on self-reported medical histories, which in-

creases the possibility of misclassification. We speculate thatcases would be more likely to report the disease than controls,

however, implying a conservative bias to the misclassification.The possibility that age of EBV infection modifies the virus’putative carcinogenic effects has been previously suggested byMelbye ci a!. (36). who found evidence ofearlier infection withEBV among children from the high-incidence Eskimo popula-

tion of Greenland than among children from the low-incidence

Danish population. It would also be consistent with the in-creased risk of NPC among persons with less education orlower socioeconomic status (who are presumably at higher riskof early infection), as found in the current study and at least one

other (23). In the current study the proportion of subjects

reporting being diagnosed with infectious mononucleosis in-

creased from 0% to 4% to 7% in persons with educationallevels of less than high school, high school, and more than high

school respectively. The higher mean age among cases thancontrols does not totally fit with this hypothesis, although itmight suggest that primary infection only in early childhood isrelated to increased risk. Also inconsistent is the preponderanceof nonkeratinizing and undifferentiated tumors (histologiesmost strongly linked with EBV) among the cases reportinghaving had infectious mononucleosis.

The inverse association between NPC and educationallevel that we observed to be consistent across histological types

might imply an underascertainment of controls with lowereducational levels from the general population. Alternatively, itmay indicate the existence of additional risk factors (e.g., earlyinfection with EBV as discussed above, diet or occupational

factors) that are correlated with education.The present study found little indication of increased NPC

risk associated with previous diagnosis of nasal polyps, sinus-itis, or repeated nose bleeds. Two previous studies, primarily

among Chinese, suggested an increased risk with a history of

chronic ear and nose diseases (24, 37). A previous reportsuggested that long-term use of nose drops or nasal spraysincreased the risk of sinonasal cancer, raising the possibilitythat the adjacent nasopharyngeal epithelium was also at risk(38). We found no evidence of an association with NPC, with

the possible exception of lymphomas. It should be noted thatthe likely inaccuracies involved in the reporting of previous

medical conditions or treatments, combined with the smallnumbers of exposed persons, suggest that the current study hadvery low power to detect such relationships.

We found associations with cigarette smoking and heavy

alcohol intake that were specific for histological type and age atdiagnosis of NPC. Future studies of this cancer need to take intoaccount histology and age in evaluating these and presumablyother environmental and genetic risk factors. The present reportsuggests that more than two-thirds of the cases of differentiated




squamous cell carcinomas arising in persons over 50 years oldcan be accounted for by cigarettes and alcohol, but leaves

unexplained NPC arising in the young, and carcinomas ofundifferentiated or nonkeratinizing histology. The finding thatinfection with the EBV later in life may be associated with

decreased risk is intriguing, although it is based on smallnumbers. This issue might be best studied prospectively usingbiological markers of infection.

References

I . Hildesheim, A.. and Levine, P. H. Etiology of nasopharyngeal carcinoma: A

review. Epidemiol. Rev., 15: 466-485. 1993.

2. Burt, R. D., Vaughan. T. L.. and McKnight. B. Descriptive epidemiology and

survival analysis of nasopharyngeal carcinoma in the United States. Int. J. Cancer.

52: 549-556, 1992.

3. Pearson, G. R. Epstein-Barr virus and nasopharyngeal carcinoma. J. Cell.

Biochem. Suppl.. 17F: 150-154, 1993.

4. Liebowitz, D. Nasopharyngeal carcinoma: the Epstein-Barr virus association.

Semin. Oncol., 21: 376-381, 1994.

5. Levine, P. H., Pocinki, A. G., Madigan. P., and Bale. S. Familial nasopha-ryngeal carcinoma in patients who are not Chinese. Cancer (Phila.). 70: 1024-

1029, 1992.

6. Lu, S. J.. Day. N. E., Degos, L., Lepage. V., Wang. P. C., Chan. S. H.. Simons.

M.. McKnight. B.. Easton, D.. Zeng. Y., and dc-The. G. Linkage of a nasopha-ryngeal carcinoma susceptibility locus to the ELLA region. Nature (Lond.). 346:

470-471, 1990.

7. Burt, R. D., Vaughan. T. L., Nisperos, B., Swanson, G. M., and Berwick, M.

A protective association between the HLA-A2 antigen and nasopharyngeal car-

cinoma in US Caucasians. Int. J. Cancer, 56: 465-467, 1994.

8. Jeannel. D.. Ghnassia, M., Hubert, A., Sancho-Gamier, H.. Eschwege, F.,

Crognier. E.. and dc-The. G. Increased risk of nasopharyngeal carcinoma amongmales of French origin born in Maghreb (North Africa). Int. J. Cancer. 54:

536-539. 1993.

9. Muir, C. S. Nasopharyngeal carcinoma in non-Chinese populations with

special reference to South-east Asia and Africa. Int. J. Cancer, 8: 351-363, 1971.

10. Henderson, B. E., and Louie, E. Discussion ofrisk factors for nasopharyngeal

carcinoma. lit: G. dc-The and Y. Ito (eds.) Nasopharyngeal Carcinoma: Etiology

and Control. pp. 25 1-260. IARC Scientific Publ. No. 20. Lyon. France: IARC.

I 978.

I 1 . Buell. P. The effect of migration on the risk of nasopharyngeal cancer among

Chinese. Cancer Res., 34: 1 189-1 191, 1974.

12. Zippin. C.. Tekawa. I. S., and Bragg. K. U. Studies on heredity and envi-

ronment in cancer of the nasopharynx. J. NatI. Cancer Inst., 29: 483-490, 1962.

13. Huang. X.. Wang. Z.. Luo, F., Hu, M., Gu, X., Gu, S., and Chen, H.Epidemiologic investigation of nasopharyngeal cancer in migrants. Chinese Med.

J.. 95: 757-761, 1982.

14. Yu, M. C. Nasopharyngeal Carcinoma: Epidemiology and Dietary Factors.Vol. 105, pp. 39-47. IARC Scientific PubI. Lyon. France: IARC, 1991.

15. Waksberg. J. Sampling methods for random digit dialing. J. Am. Stat. Assoc.,

73: 40-46, 1978.

16. Breslow, N. E., and Day. N. E., Statistical methods in cancer research. The

Analysis of Case-control Studies. Vol. 1 . IARC Scientific PubI. No. 32. Lyon,

France: IARC, l98t).

17. Statistics and Epidemiology Research Corporation. EGRET Statistical Soft-

ware. Seattle, WA: SERC, Inc.. 1995.

18. Bruzzi. P., Green, S. B., Byar. D. B., Brinton, L. A.. and Shairer, C.Estimating population attributable risk for multiple risk factors using case-control

data. Am. J. Epidemiol.. 122: 904-914, 1985.

19. Shanmugaratnam. K.. and Sobin. L. Histological typing of upper respiratorytract tumors. Intemational Histological Typing of Tumors, No. 19, 32-33. Ge-

neva: WHO. 1978.

20. Giannoudis, A.. Ergazaki. M.. Segas, J.. Giotakis. J.. Adamopoulus. G.,

Gorgoulis, V.. and Spandidos. A. Detection of Epstein-Barr virus and humanpapillomavirus in nasopharyngeal carcinoma by the polymerase chain reaction

technique. Cancer Lett.. 89: 177-181, 1995.

21. Doll. R.. Peto. R.. Wheatley. K.. Gray. R.. and Sutherland, I. Mortality in

relation to smoking: 40 years’ observations on male British doctors. Br. Med. J..

309:901-911,1994.

22. Hanley, A. J. G.. Choi, B. C. K.. and Holowaty, E. J. Cancer mortality among

Chinese migrants: a review. Int. J. Epidemiol., 24: 255-265. 1995.

23. Chen, C. J., Liang, K. Y.. Chang. Y. S,. Wang, Y. F.. Hsieh, T., Hsu, M. M..Chen, J. Y., and Liu. M. Y. Multiple risk factors of nasopharyngeal carcinoma:

Epstein-Barr virus. malarial infection, cigarette smoking and familial tendency.

Anticancer Res., 10: 547-553, 1990.

24. Yu, M. C.. Garabrant, D. H., Huang. T. B.. and Henderson. B. E. Occupa-

tional and other non-dietary risk factors for nasopharyngeal carcinoma in Guang-

zhou, China. Int. J. of Cancer, 45: 1033-1039, 1990.

25. Chow, W. H., McLaughlin. J. K.. Hrubec. Z.. Nam. J. M., and Blot. W. J.Tobacco use and na.sophaiyngeal carcinoma in a cohort of U.S. veterans. Int. J.

Cancer, 55: 538-540. 1993.

26. West, S.. Hildesheim. A., and Dosemeci, M. Non-viral risk factors for

nasopharyngeal carcinoma in The Phillipines: results from a case-control study.Int. J. Cancer. 55: 722-727. 1993.

27. Nam. J. M., McLaughlin. J. K.. and Blot, W. J. Cigarette smoking. alcohol.

and nasopharyngeal carcinoma: a case-control study among U.S. whites. J. Natl.

Cancer Inst., 84: 619-622. 1992.

28. Mabuchi, K.. Bross, D. S., and Kessler II, I. I. Cigarette smoking and

nasopharyngeal carcinoma. Cancer (Phila.), 55: 2874-2876, 1985.

29. Blot, W. J. Alcohol and cancer. Cancer Res., 52: 21 l9s-2123s, 1992.

30. International Agency for Research on Cancer. Alcohol drinking. IARCMonographs on the Evaluation of the Carcinogenic Risk to Humans, Vol. 44.

Lyon. France: IARC. I 988.

3 1 . Swann, P. F. Effect of ethanol on nitrosamine metabolism and distribution.Implications for the role of nitrosamines in human cancer and for the influence of

alcohol consumption on cancer incidence. In: 1. K. O’Neill. B. C. Von Borstel. C.T. Miller, J. Long. and H. Bartsch (eds.). N-Nitroso Compounds: Occurrence.Biologic Effects and Relevance to Human Cancer. IARC Scientific PubI. No. 57.

Lyon. France: IARC, 1984.

32. Yu, M. C., Nichols, P. W., Zou. X. N.. Estes, J., and Henderson, B. E.

Induction of malignant nasal cavity tumors in Wistar rats fed Chinese salted fish.

Br. J. Cancer. 60: 198 -201 . I 989.

33. Zou. X. N.. Lu. S-H., and Liu, B. Volatile N-nitrosamines and their precursorsin Chinese salted fish-a possible etiological factor for NPC in China. Int. J.

Cancer, 59: 155-158. 1994.

34. Yu, M. C.. Mo, C. C., Chong. W. X., Yeh, F. S., and Henderson, B. E.

Preserved foods and nasopharyngeal carcinoma: a case-control study in Guangxi.China. Cancer Res., 48: 1954-1959, 1988.

35. Deleyiannis. F. W. B.. Thomas, D. B., and Vaughan, 1. L. Alcoholism and

abstinence as independent predictors of survival in patients with head and neck

cancer. J. NatI. Cancer Inst., 88: 542-549, 1996.

36. Melbye. M., Ebbesen, P.. Levine. P. H., and Bennike, 1. Early primary

infection and high Epstein-Barr virus antibody titers in Greenland Eskimos at

high risk for nasopharyngeal carcinoma. Int. J. Cancer. 34: 619-623, 1984.

37. Henderson, B. E.. Louie, E., Jing, J. S., Buell, P., and Gardner, M. B. Risk

factors associated with nasopharyngeal carcinoma. N. EngI. J. Med., 295: I 101-

1106, 1976.

38. Strader, C. H., Vaughan. T. L., and Stergachis, A. Use of nasal preparations

and the incidence of sinonasal cancer. J. Epidemiol. Community Health, 42:

243-248, 1988.



1996;5:587-593. Cancer Epidemiol Biomarkers Prev T L Vaughan, J A Shapiro, R D Burt, et al. factors by histological type.Nasopharyngeal cancer in a low-risk population: defining risk

Updated version

http://cebp.aacrjournals.org/content/5/8/587

Access the most recent version of this article at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cebp.aacrjournals.org/content/5/8/587To request permission to re-use all or part of this article, use this link



http://cebp.aacrjournals.org/cgi/alerts

mailto:[email protected]



Documents

Nasopharyngeal Cancer in a Low-Risk Population: Defining ...Nasopharyngeal cancer (NPC) is a major public health problem in parts of Southeast Asia and North Africa, but is rare among