) ¿(a) ic~i1367

h tti.~~rj1~’tt.y li~

l(’~ti.!~[,y rio~iyI,~fii

¿iit~r1~tL~-;.(:c.~h’)

Smoking Tobacco, Oral Snuff, and Alcohol in theEtiology of Squamous Cell Carcinoma of theHead and NeckA Population-Based Case—Referent Study in Sweden

Freddi Lewin, M.D.”2

Statfan E. Norell, M.D.’

Hemming Johansson, B.SC.1

Per Gustavsson, M.D.3

Johan Wennerberg, M.D.4

Anders Biörklund, M.D.4

Lars Erik Rutqvist, M.D.’

Oncologic Centre. Karohnska Hospital. Stock-hoirn. Sweden

Department of Oncology. Huddinge UniversityHospital. Huddinge. Sweden.

Departriient of Occupational Health. KarolinskaHospital. Stockholm. Sweden.

‘Department of Otorhunotaryngology’Head andNeck Surgery. University Hospital. Lund. Swe-den

Presented in part at the Second World Congresson Laryngeal Cancer. Sydney. Australia. ~ebru-ary 20—24. 1994

Supported by grants from the Cancer Societyof Stockholm. the Swedish Cancer Fund. andthe research fund of the Swedish Tobacco Com-pany

The authors acknowledge Karin Andersson.R N.. and Ulla Kltnga. R N.. for so carefutfy per-forming the interviews They also thank loomSingnomklao for managing the data

Address for reprints Freddi Lewin. M O Onco~logic Centre, Karolinska Hospital. S-ill 76Stockholm. Sweden.

Received June 23. 1997. revision received Octo-ber 7. 1997: accepted October 7, 1997

BACKGROUND. This case-referent study was conducted to elucidate the rote ofselected exogenous agents in the etiology of head and neck cancer. The factorsstudied were tobacco smoking. alcohol intake, the use of moist oral snuff, dietaryfactors, occupational exposures. and oral hygiene. in this first report, the authorsdiscuss the impact oftobacco smoking, the useof oral snuff, and alcohol consunip.

lion.METHODS. The study base was approximately 2 million person-years at risk andconsisted of Swedish niales age 40—79 years living in 2 geographic regions duringthe years 1988-1990. A total of 605 cases were identified in the base, and 756controls were selected by stratified random sampling from population registries

covering the base.RESULTS. Among those who were tobacco smokers at the time of the study, therelative risk of head arid neck cancer was 6.5% (95% confidence interval, 4.4—9.5%I. After cessation of smoking. the risk gradually declined, and no excess riskwas found after 20 years. The relative risk associated with alcohol consumption ofSt) grams or more’ per day versus less than 10 grams per day was 5.5% (95%

conlide~iiceinterval, 3.1 —9.6%l. An almost multiplicative effect was found for to-bacco smoking and alcohol consumption.CONCLUSIONS. 1 nhacco smoking and alcohol intake liad a strong interactive effecton the risk of squamous cell carcinoma of the head and neck. Moderate alcohol

intake (10—19 grants per day) had little or no effect among nonsmokers. No in-creased risk was found (or the use of Swedish oral snuff. Cancer I998;82~1367—75. e I.9~Uf.4us,’rwapi Cancer .Soclett’.

KEYWORDS: tobacco smoking, alcohol, oral snuff, squamous cell carcinoma, headand neck cancer, esophageal cancer, case—referent study, epidemiology.

There is a geographic variation in the incidence of cancer of thehead and neck among different countries of the world and among

difFerent regions within a country.’ This indicates that environmentalFactors mayplay an lnipouan role in the pathogenesisof cancer of thehead and neck. In Europe. Sweden has one of the lowest incidences ofthese cancers. In Sweden, the majority (two-thirds) of patients withsquanmus ccli carcinoma of the head and neck are males. It has beenshown previously that tobacco smoking and alcohol intake are majorrisk factors.~In Asia. chewing-tobacco causes a high incidence oforal cancers,~and in the U.S. there have been reports of oral snuff asa risk factor in oral cancer.sc About 15% of all adult males in Swedenuse, or have used for part of their lives, an oral snuff produced

1998 American Cancer Society

1368 CANCER April 1, 1998 IVolume 82 I Number 7

mainly in Swedeti. ilie snuff is a moist, nonfermentedtobacco product originat¡ng from the species Nicoli-nurn ial,acu,n. It is placed on the gum under the upperlip and is rarely used outside the Scandinavian coun-tries he ‘Swedish’’ moist oral snuff is knownto causereversible white J)atCheS 00 the site of application.8

lhe possible association of Swedish oral snuff withcancer of the head and neck has not yet been deter.tuilled. 1 Iowcvcr, due~to pott’ntiahty carcinogenic sub-stances contained in Swedish oral snuff, a concernregarding such a possible association exists.

lo identify possible factors involved in the etiol-ogv of cancer of the head and neck (oral cavity, oro-aocI hvpopharvnx, larynx, and the esophagus) amongmen. a population-L)ased case— referent studywas per-formed over 3 ~‘earsin 2 defined geographic regions inSweden. lobacco smoking. oral snuff, alcohol, dietaryfactors, occe~patioiial(-xf)osures, and indicators of oralItvgiette were itwt’stigatcd. In this lirsl report, we dis-cuss the impact of tobacco smoking. tite use of oralstiull. and alcohol cotlstIn)ption.

MATERIALS AND METHODSlite’ stud base’ was the’ iwr~~n-t¡inc generated by allnit-I) horn in Sweden. ages -ID— 79 ~ living in (andincluded in the population registers of) tile Stockholmcounty or tite southern I,e’althicare region of Swedeitduring mitt’ stetdv period January 1988 through JanuaryI 991. .1 Ints. 11w stud base’ was approximateb’2 millionpersi-y~i~ at risk.

CasesEfforts were made to identify all incident cases of can-cer of tite l,e’ad amid neck (squamous cell carcinomaof the oral cavity, tiro— atid hvpophiarvnx. larynx, andt’so~)h)agus)that occurred itt the stud base’. Cancersuccurt lug otttsidt’ tite’ study liase’ were not included..1 In’ cases were identified at weekly multidisciplinaryceiiife’rt’nce’s at :ili of tite’ six ear, nose, and throat de’-~ tuents IFNI) where’ almost all head and neck cati—cers Iii the two regions we’re treated. In addition, pa-itents ~vithiesophageal cancers nut diagnosed at theLXI departments we’re reported 1w all the th’part-itients of surgery ill the two regions. lo identify casm.’snot presented at the conferences or reponed from tite’(1t’J)arI!iit’titS of surgery. information was obtained cv—cry second week on the recent reports to lht’ regionalcancer registers in Stockholm ¿itid in the’ southern re’-gion. About 10% of tite case’s we’re idcniilit’d ill thisway. Notification of the regiotial cancer registries,lIboLIt new cancer cases is compulsory for 1)011) t’liiii -

cians aocI pathologists. Of all caticers nl the head andfleck, almost 99’i~,arc being rt’gisie’red in tIte regional

cancer registries.9 Cancers identified incidentally atautopsy were not included.

ReferentsThe referents were selected by stratified random sam-pling every 6 months during the study period from acomputerized population register in each region.Stratification was by region (Stockholm and the south-ern region) and age (40—54, 55—64, and 65—79 years).The population registers are updated every month.

InterviewsCases and referents were asked to participate in aninterview on life-style and environment. Informedconsent was obtained from each subject. The inter-view followed a structured questionnaire coveringsmoking history, the use of oral snuff, and alcoholintake as well as dietary factors (food frequencies) andindicatorsof oral hygiene (the numberof toothbrushesused per year and the neiniber of visits to a dentist peryear) and occupational exposure. All interviews wereconducted by two nurses, one in each of the two re-gions. The nurses were trained for health interviewsand for treating cases and referents alike. Most of thecases were interviewed at the hospital. The cases wereinterviewed approximately I month after diagnosis.The delay was deliberate, to give the patients time toget used to the new medical situation. Referents wereusually Interviewed in their honies.

Smokinglifetime smoking histories included information onthe time when a subject began or stopped smokingand the average number of cigarettes, ciganillos, cigars,and grams of pipe tobacco smoked per day duringdifferent time periods. Total consumption of smokingtobacco was calculated by adding the quantity of to-bacco smoked dttringdifferent time periods, consider-ing I cigarette or cigarillo equivalent to I gram atidI cigar equivalent to 5 grams of tobacco. The meanintensity of smoking was calculated by dividing thesubject’s total consumption by the duration of smok-ing. “Ever-smokers” were men who had ever regularlysnmked at least 7 grams of tobacco per week. To avoidthe possibility that cases would he classified as ex-smokers because they had stopped smoking dite toinsidious cancer symptoms, subjects were consideredcurrent smokers if they smoked 1 year prior to thetime of the interview.

Oral SnuffOral snuff usage was recorded in a similar way assmoking history. considering men who had ever regu-larly used I package (50 grams) per week as ever-users

r

Tobacco and Alcohol In SCCHN EtiologylLewin el al. 1369

TABLE: ITotal Number of Cases Identified, Referents Selected, NumbersInterviewed and lost, and Reasons for NonparticipationIn the lnteniews

No. 1%) of cases No. (%) of referents

tdctitificd/sclected 605 (100~4 756 (100%)list not inirrv~ewecf~ 60(10%) hIS 115%)

li~utised 17 80l)i~ahIed 9 8Dead 30 6Not located — 21Other rea~on~ 4 —

tnlrr~leo rd 54 )90%l 64) (85%)

and men who used oral snuff 1 year prior to the timeof the interview as current users. Total consumption,duration, and mean intensity of usage were calculatedin the same ways as for smoking tobacco.

Alcohol IntakeIntake’ of alcoholic beverages 5 years prior to the timeof the’ inte’re-k’w was assessed using a questionnaireslightly iiiocli(Ie’d from Gerhardsson de Verdier et al.”’It provi(le’cl information on the intake of beer, wine,atir.h hard liquor, using seven categories of consump-tino frequency. and the average amount consumed oneach occasion. ‘Ibis information was translated intograins of alcohol per day using a data base at the Swed-¡sI u Nut k mal I noel Ad in in ist ration -

Data Analysis1 he refere’nts ~ se’lectecl to provide information otiexposure’ fre’que’nc-ies in the person-time that gener-ated tite’ case’s. ‘11w relative’ risk (RH. incidence rateratio) was cale-ulated by logistic regression anal sis)~.tdjttcttnents we’re’ made according to study design forage’ tthre’t’ categories) amid region (two categories). 10some analyse-s. adjustine’nts were also made for to-bacco smoking (current smokers. ex-smokers. andthose- svhio never smoked) or alcohol intake 1< 10, 10—19. ~()— 49, and 50 grams per das’). or both. As a checkof residual confounding. adjustments were also madefor age’ in 5-year categories, duration of smoking. oralitvgiene’. aoci certain dietary factors.The EGRET (1988)cornpule’r program from the Statistics and Epidemiol-ogy hte’st’are-h Curpor~mtionwas used to process thedata.

RESULTSA total of 605 case’s we’re identified, and 756 referentswe’re~se’le’ctecl. Ninety percent of the cases and 85% oftIte’ referents participated in the interviews. Reasonsfor not participating are shown in lable I

The effect of smoking was similar when men whosmoked cigarettes only (RR 3.7, 95% confidence in-terval fClJ = 2.5—5-5) were compared with those whosmoked cigarillos. cigars, or a pipe (RR 4.), 95%Cl = 2.3—7.4) and those who mixed use of differentsmoking tobacco (BR = 4.1, 95% Cl = 2.8—6.1). Dueto there being only two pure cigar smokers in the ma-terial, it was not possible to perform the RR analysisfor this subgroup. Allsmoking tobacco was consideredtogether in Table 2, showing the RRs associated withdifferent aspects of smoking. The risk was consider-ably lower for ex-smokers than for current smokersand was related to time since smoking cessation. Noincreased risk was found for men who had stoppedsmoking more than 20 years previously. There was alsosome association between risk and the mean intensityof smoking. Cessation of smoking was, however, morecommon among men who smoked only a few ciga-rettes or grams of tobacco per day than among menwith a high daily consumption. To evaluate the impactof the mean intensity of smoking aside from smokingcessation, it was investigated in current smokers: BR= 6.1 (95% CI = 4.0—9.5) for men smoking <15 gramsper day. BR = 6.1 (95% Cl = 4.0—9.3) for men smoking15—24 grants per day, and RR = 6.6 (95% Cl 3.4—12.7) for men smoking ~25 grams per day. This sug-gested that aside from the effect of smoking cessation,there was little or no impact of mean smoking inten-sity, If so, the impact of total consumption would es-sentially reflect an effect of duration of smoking. be-cause total consumption equalled mean intensitymultiplied by duration of smoking. As shown in Table3, smoking cessation and the duration of smoking eachhad a decisive impact on risk.

The cancer subsites in the interviewed cases were:the oral cavity in 128 cases, the pharynx in 138 cases(75 oropharvnx and 63 hypopharynx), the larynx(mainly glouic) in 157 cases, and the esophagus in123 cases. Analysis by cancer subsite showed similarresults, although the relative effect of smoking wasntore pronounced for cancers of the pharynx and lar-ynx than for cancers of the other subsites. For currentsmokers, the RRs (with 95% Cls) were as follows: forcancer of the pharynx, RH 8.5 (4.0—18.2); larynx~RR= 7.5 (3.9—14.2); esophagus, BR 5.2 (2.6—10.3); andoral cavity, RH = 4.9 (2.6—9.2). For roen who hadsmoked 45 years or longer: pharynx, RB 10.1 (4.6—22.1); larynx, BR 7.6 (3.9— 14.7); esophagus, BR 5.4(2.7—11.0); and oral cavity, KR = 6.3 (3.2—12.4).

Overall, the use of oral snuff had little or no effecton risk, as shown in Table 4. In an analysis performedwith the reference category “never-tobacco-users,” preci-sion was very low, as there were only 9 cases arid 10referents who had ever used snuff but had never smoked

1370 CANCER April 1, 1998 / Volume 82 / Number 7

1’ABI.F. 2Smoking~and Relative Risk ofElead and Neck Cancerk In SwedIsh Men Ages 40—79 Years

Relative rIsk (95% confidence Interval)adjusted for

Destgn~+Smoking No. or cases No. of referents Design’ alcohold

Never stiroked 44 193 1.0 1.0Ever srrrokr-d SOl 448 5.0 (3.5—7.0) 4.0 (28-571current snroker~ 385 214 8.4 (5.8—12.2) 6.5 14.4—95)I.o-s;rioker~ It6 234 2.1 (t.4—3.1) 1.9 (1.3—2.8)Stopped smoking

1-10 vrs ago 6) 75 3.5 (2.2—5.7) 32 (2.0—52)11-20 rs ago 32 76 1.8 (1.1—3.11 1.7 11.0—2.9)-2( vr~ago 23 83 1.1 10.6-2.0) 0.9 (O.5—I.7j

Age at starrctS vr~ (lO 7? 6.5 (4.2—10.1) 5.0 (3.2—7.9)IS-IS vrs 257 220 5.2 13.6-7.6) 4.0 (2.7—5.9)20-24 vr~ lOt 102 4.4 (2.8—6.7) 3.8 (2.4-5.9)~-25‘,r~ 33 49 2.8 (1.6—4.9) 2.6 (1.5—4.6)

lhrr,ilrorr of smoking-30 vr’, 50 15� tI (0.8—2.0) 1.2 (0.7—1.9)30-44 vr~ 168 148 4.9 (3.3-7.3) 3.9 12.6-5.9)~-45sr~ 283 141 9.3 (6.3—13.8) 7.2 (4 8—10.8)

lora) concunrprron’c l2’~kg tobacco Si 145 1.6 11.0—2.5) t.5 (1.0—2.4)t25-2i’i kg mohaceri 18) 146 5.5 (3.7—82) 43(2.9—6.5)‘250 kg tr’hacei.r 287 IS? 7.5 (5.t-1I.0) 5.9 (4.0—8.8)

lnterr~ir~rif ~rrroking’‘ .

-. IS g tobaccoda~ 202 211 4.1 (2.8—6.0) 3.4 (2.3-5.!)¡5-24 g tobacco’

d,i~ 2t0 189 S.S (38—8.1) 4.4 (2.9-6.5)-2” g tobacco das 69 48 6.5 (4.0-10.7) 4.8 (2.9-St)

livep rnha)rr~’341 17� 8.9 (6.1-13.0) 6.? (4.5-tOO)

No It 33 5.3 (3.0—9.3) 3.9 (2,1—7.0)

,~.,r.ik,.~ (I.~

s.4u~niu.cr-li carun.nia c~he nr.,I c~rtior., anJ h ipharrnr. arrio. an.J e’arphagvc‘er- ‘I’ - ~4 ~ ~‘ ‘0 ns ari.l r,’~.e~,ctJr,Imand h~‘.ouih Ss,eden healthcarr arr~a5.s~’cac~rr,es. O ti- I’. ~ - e rkuh,~dj,

ii ci~.ir~I I I .l~:..’

I.”., ,.r..oopi.n d,,is1r-r~1’. duO~,’c,,f ~ ~p,’r Ja.- 5’~ •,, t ~ .n. Icr’ ILEj ru..:i~I’’, ta’.—. ¿rl.,’ ~‘ rrIrrr-nIs

r

tobacco. llw HR (95% Cl) for ever-users of snuff was RH4.7 (1.6- 13.81. For current users. RH 3.3 (95% Cl =

(1.8— 12.0), and (cur ex-uSers. KR = 10.5 (95% Cl = 1.4—117.8), further illustrating the low precision. When fannersotokers were the reference category. the precision washighte’r. with 24 cases and 46 referents who had ever usedsnufT. For ever-users of snuff withi this reference category.HIt 1.1(95% (:1 = 0.6— 1.9). For cunenl users of snuff.HR 1.4 (95% Cl 0.7—2.8), and for ex-users RH = 0.8(95% (:1 = 0.4—1.8). With current smokers as a referencecategory, the [(Rs (95% GIs) were as follows: RH = 0.8(0.5— 1.2) for ever-users of snuff. HR = 0.6 (0.3—LI) forcurrent snuff users, artcl HR 1.0 (0.5—2.0) for ex-users.

In an analysis by anatomic subsite, precision was againlcn~’.l’he ROts by subsites are shown in Table 5.

The effect of alcohol intake is illustrated in Table6. The results suggest a gradual increase in the risk ofcancer of the head and neck with increasing alcoholintake. However, moderate alcohol intake (10—19grants per day) had little or no impact on the risk ofcancer in ex-smokers and in men who never smoked(Table 7). Moderate alcohol consumption was foundto increase the risk only among current smokers. Thejoint effect of a high alcohol intake (�20 grams perday). ~sithan HR of 4.2 and current smoking RR = 6.3,was nearly multiplicative; BR = 22.1,

Tobacco and Alcohol In SCCHN Etlology/Lewin el al. 1371

TABLE 3Duration oF Smoking for Current Smokers and F.x-SmokersandRelative Risk of Head and Neck Cancer In Swedish MenAges 40-79 Years

fluration of smoking

Relative risk (95% confidence Interval)No. of eiposed cases/exposed rekrents

Current smokers Li-smokers

.-4S yr. 7.3 (4.8-11.0) 4.4 (2.4—8.0)24711t3 36/31

50-41vr. 61(3.8-9.8) 2.4 (1.5—4.0)¡20/74 48/74

<30yr. 2.4 (1.1—5.3) 1.0 (0.6—1.7)t8/27 32/129

tImo~iosr’d er-net smol.ec’,: 4) cacect93 reír-tenislic)aiisc n.b arr adjuct,-d for age 143-54.55-64.65-70 ~‘t~I.region ISIOCkhOIn. andthe South SwedenIst-alrhca.r- arta and alcohol imale -rlOlO-19.2(-49. wSO alcoholtdav),

Analysis by subsite showed the strongest relativeefft’ci of alcohol for cancer of the esophagus (RH8.6, 95% Cl r 3.8—19.2) and pharynx (RR = 8.5, 95%(:1 4.0— 18.1) at an alcohol intake of ~50 grams per(lay. lar the other stibsites, the corresponding effectswelt- as follows: oral cavity. RH 5.7 (95% Cl = 2.8—I 1.9) and hars’nx, RH = 2.0 (95% Ci = 0.9—4.7). The[(Rs were adjusted for design and smoking, as shownin l ablc- 5.

lo check for residual confoLinding, RRs for smok-(tug, oral snuff, and alcohol (Tables 2, 4, and 5) wereadjusted lar age iii 5-year categories, and RRs for snuffand alcohol (l’ables 4 aoci 5) were also adjusted forclur.ution of smoking. This, however, heft the resultsvirtually unchartged. In addition, adjustments for di-t’tarv intake of calories, protein, fat, carbohydrates, fi-hers, and vitamins and for indicators of oral hygieneliad little or no impact on these results.

DISCUSSIONl lic’ possibility of bias due to identification of casesand selection of referents is important in any case—referent swcl~-.lo avoid such bias in thecurrent sudy,we made efforts to identify all incident cases of headand tteck cancer that occurred during the studyperiodin a population defined by age. gender. and residence.‘lhtis was facilitated by close cooperation with the clini-cians imtvolved and by the availability of population-based cancer registries, in addition, the referents wereselected front continuously updated registers of theliase population for the purpose of obtaining a repre-sentative sample of the person-time that generated thecases.

l’hc- inter\-iesvs were completed by 90% of thecases identified amid by 85% of the referents selected.

Thus, even a substantial difference in exposure be-tween those interviewed and those not interviewedwould only have changed our results modestly. Toavoid differential misclassification, the interviewerswere trained to ask the questions in such a way thatany impact of the disease on the answers would beminimized. However, for medical reasons, there weredifferences in the interviews. The cases were mostlyinterviewed at the hospital and the referents were usu-ally interviewed in their homes. The cases were inter-viewed about I month after diagnosis but had experi-enced symptoms for some time prior to diagnosis.Some cases could have reduced (or increased) theirtobacco or alcohol consumption due to such symp-toms. To avoid this source of bias, exposure informa-tion for cases and referents did not include smokingand oral snuff usage during the last year prior to theinterview, and information was obtained on alcoholintake 5 years prior to the interview. However, smokerswho have reduced their smoking tend to underreporttheir past smoking.ma Thus, if our cases had reducedtheir smoking due to symptoms of disease, smokingcases could have underestimated the number of ciga-rettes they smoked per day in the past. This wouldresult in some underestimation of the effect of meanintensity of smoking in the current study. Underre-porting of alcohol intake is another possibility. Pa-tients with a serious disease (our cases) could be lesslikely to underreport their alcohol intake than healthysubjects (our referents). If exposed referents were clas-sified as unexposed, the effect of alcohol intake wouldbe overestimated. If highly exposed referents wereclassified as moderately exposed, the effect of a highalcohol intake would also be overestimated, but theeffect of moderate alcohol intake would be underesti-mated.

Tobacco smoking has previously been shown toincrease the risk of several cancers, including squa-mnous cell carcinoma of the head and neck. We founda fourfold increased risk for ever-users of smoking to-bacco (HR = 4.0, 95% Cl = 2.8—5.7) for all sites. l’hiswas well in accordance with the results of other stud-ies.’4’~Mean intensity of smoking had little or noimpact, but the risk increased with the duration ofsmoking. Similarly, Roth man et al. found only a minordifference in risk according to mean intensity, whereasBrugere et al. found a strong correlation between in-tensity and risk.’7’0’ However, whether the intensityindicated is the mean intensity or current intensity isunclear. In results similar to those of our study, Blotet al. found a smaller difference in risk regarding inten-sity of smoking as compared with duration of thehabit. Their risk estimates were generally lower thanours.’4-’5 Bundgaard et al. found that the risk of oral

7 .4

1372 CANCER April 1, 1998 I Volume 82 (Number 7

TABLE 4Oral Snuff Usage and Relative Risk of Itead and Neck Cancer In Swedish Men Ages 40—79 Years

Relative rIsk (95% confldencehiten-al) adjusted for

Oral snuff usage No. of cases No. of referents DesIgn’ DesIgn’ + alcohol5 and smoking’

Never used 462 550 1.0 1.0Ever used 83 91 1.1 (0.8—15) 1.1 (0.7—1.5)Current users 43 50 1.0 (0.7—1.6) ¡.0(0.6-1.6)h-users 40 II ¡.2 (0.8-1.9) 1.2 (0.7—1.9)Age at start

<25 vn 39 43 1.1 (0.7—1.7) 1.0 (0.6-1.6)o’75 yrs 41 48 1.1 (8.7—1.71 1.1 (0.7-1.8)

hluraiion of usage<30 vms 52 59 1.1 (0.8—1.7) 1.0 (0.7—1.6)�30vrs 31 32 1.1 (0.7—1.9) 1.1 (0.6-2.0)

lota! concurrtption<125 kg 5 63 1.1 (0.8-1.7) 1.0(0.7—1.6)?t25Eg 26 28 1.1 (0.6—1.9) 1.1 (0.6—2.0)

Irnencit~of ucagea~50gweck 45 57 1.0(0.6-1.4) 0.8 (0.5—1.3)>50 g/wet’k 38 34 ¡.4 (0.9-2.3) 1.6 (0.9—2.6)

A~t45-54. 55-54. ~ orn- and regon StucI.holrn and rSe South Swedeur healthcaie ateal.‘Four cat~-gu~c’.--:10. ¡0.15. 75.49. aSO g alcohol 4.19:lhree caIq.’.n,r. rre,ersrr,ohrrs n-se,,,~errcunraI srr,ohers

‘101.1! con.unpoorr drsrde.1 b~duration rif os.tgr- per weeL

1ARLE 5Oral Snuff Usageand Relative Risk of Head and Neck Cancer In Swedish Men Ages 40—79 Years by SIte

Oral rasity Lar~nn Esophagus Pharynu

Oral anulO usage Cacerjrelerents Rl!’ Cases/reIerenia RR~ Cases/referents Rl!’ Casesireferents Rl!’

‘seso u’e’d ¡03550 ¡.0 133)550 1.0 103/550 1.0 123/550 1.0used 259) ¡.4 08.2.4) 24/91 0.903-I.5l 19/91 1.2 (0.7—2.2) 15/91 0.7 (0.4—1.3)

(.urrerlt users ¡050 ¡.0 10.5-2.2) I5’50 1.0 0.5-1.9) ¡0/50 1.1 (0.5-2.4) 8/50 0.7 (0.3—1.5)I sisen 151t ¡809-3 9~l 0.8 )0.4-l,7t 9141 1.3 0.6—3.1) 7/41 0.8 (0.3—1.9)

h,)..~,,,,s.i Kit ‘$5’. c,,nr,J,’n,e urtrrnai. ar~aJ’a’a’d Kr ~r ~ ~ ~ ~ ~w xlJsoZ.m and the South Sweden heulr}scare *:eat.atrlohirrt ineorc ~ir,o5ers,en-caroLers, current sniokers)..a-~j~l.’h’ nt~kr - ir to. Is 2 4-, _~ ak/si da,

cancer increased with current daily consumption ofsmoking tobacco. but also with lifetime consump-tion)°luyas et al.. however, found that the risk in-creased with nican intensity. ~“ For supraglottic cancerthey found a 1111 of 2.8 195% Cl = 1.2—6.8) for a meanconsumption of 1—7 cigarettes per day and a RH of24.0 (95% Ci 11.8—48.7) for a mean consumption ofmore titan 26 cigarettes per da>’. We found a tendencyiowards a higher RH for deep inhaling of tobaccosmoke. luis was also reported by l’u>’mts et aL, butonly for glottic cancer.’4 It is known that the risk ofdeveloping cancer of the head and neck decreasesafter smoking cessation. Blot et al. and l’uyns eL al.

found no excess risk after 10 years,’4-’9 whereas Spitzet al. found no excess risk later than 15 years aftercessation of smoking.2°We found a gradual decreasein risk up to 20 years after smoking cessation in ourstudy. Some differences were found in the magnitudeof the effect for different subsites. This was also inaccordance with previous results.’8”9 1-lowever, thenumber of cases in each subsite was small, and theresults should therefore be interpreted with caution.

Alcohol intake may increase the risk of head andneck cancer, according to previous studies.24’2124

Even though alcohol per se is not mutagenic, possiblemechanisms for alcohol-related carcinogenesis have

Tobacco and Alcohol In SCCHN Etiology/Lewin et ah. 1373

TABI.F, 6Alcohol intake and Relative Risk of Head and Neck Cancer In Swedish Men Ages 40-79 Years

.

Atrohol intake No. of cases ‘

.

No. of referents

ReIstlue risk (95%adjusted for

confidence interval)

Deshgn’DesIgn’ +smoking°

<10 glday 18510-19 g:dav II?20-49g/da~ 171n,54S g/dav 72

363¡56lOE21

1.0¡.6 (1.2—2.1)3.8 (2.8-5.2)8.4 (4.9—14.3)

1.01.3 11.0—1.8)2.7 (1.9-3.8)5.5 (3.1-9.6)

‘Ag~10-54 ~ tin) and region (Stockholm and the Soulhtes cd/t.g,rit-~Jn,’n~rsrr:i~5er,eo’sniokers, curreri smokers),

th Sweden Iaealthcare area).

TABLE 7Smoking, Alcohol Intake, and Relative Risk of Head and Neck Cancerin Swedish Men Ages 4 0—79 Years

Relative rIsk (95% confidence Interval)No. of cases/no, of referentsAlcohol intake

Smoking ~-20g/daE 10-19 g/day <10 g/day

arts_ri’ oritokers 22h13 0-37.8 10.4 (5.9—18.31 6.3 (3.7—10.5)I%’62 84151 105/101

I s-nrtoikr. 5.4 2.8-1021 22 (1.2—4.1) 2.4 (1.4—4.1)3440 26/68 56/126

\t-st’t sntoked 4.2 11.8-97’ 12)0.5—3.1) 1.0¡3/2(3 7/37 24/136

it,.,:., ri-i’ ¿!e ajr.,.rrri ir 11

s 40-54 5’~5455.’9 ‘.r~and region Stockholm and the Southr,.,~:,,,Cr ¿‘e.

been discussed by 1-cato and Nomura.25 In our stud>’,Hit 5.5 (95% CI 3.1 —9.6) for consumption of morethan 50 grants per day, compared with less than lOgrams per ciav. after adjustment for smoking. A dose-dependent increased risk was found for tumors in theoral cavity. pharvux. and esophagus. These resultsseenwd to be in accordance with other studies.2’’~’28

Lvc-n titougit not completely comparable, the magni-tude of lOIts itt our study regarding alcohol consump-tion were similar to those of other studies. We founddifferent Hits for different subsites of the head andneck. This is also known from others. No significantincreased risk was found for tumors in the larynx. evenat tin’ highest dose level, suggesting a local effect ofalcohol oit the mucosa of the upper digestive tract.lii this study, supraglottic canëer was not analyzedseparately, as the patients in Sweden with supragl000iccancer represent only about 20% of all patients withlarvttgeal cancer. Others have found an increased risk

for supraglottic tumors of the larynx from alcohol con-surnption.°~9Modest consumption of alcohol only hada minor impact on the risk of head and neck cancer.Franchesi et al. got the same result in their study.2

Hedberg et al. found a significant increase of laryngealcarcinoma among alcoholics as measured by theMichigan alcoholism screening test, even after adjust-ment for cigarette smoking.29 We did not classify ourpatients as alcoholics or nonalcoholics.

Notable is the almost multiplicative effect of com-bined high exposure to both tobacco smoking and al-cohol. with an RR of 22.] (95% Cl 12.9—37.8). Thisresult has also been found by others.3-’4-’5-303t Maier etal., for example, found that heavy smoking and drink-ing together increased the risk 146 times, and Bunde-gaard et al. found a multiplicative effect (with RR =

80) for more than 20 grams tobacco smoked and morethan 5 drinks per day.3-’°The mechanism behind thepathogenesis is largely unknown. However, there iscircumstantial evidence for a genetic link to I)NA re-pair. An increase in numeric and strttctural chromo-somal rearrangements in the normal mucosa of sniok-ers compared with nonsmokers has been recorded.32

A defect in DNA repair might explain the impact ofalcohol on cancer induced by tobacco smoking. Onthe molecular level, it has been shown that the fre-quency of p53 mutations among patients with squa-mous cell carcinoma of the head and neck are higherin smokers than in nonsmokers and even higher insmokers who also drink alcohol.33 The magnitude ofthe RR associated with tobacco smoking and alcoholconsumption varies among different studies. This maypartly be due to differences in tite consumption pat-terns indifferent parts of theworld. We did not investi-gate the impact of different kinds of smoking tobacco,as the vast majority of smokers in Sweden smokeblond tobacco (dark tobacco is used by less than 1%

1374 CANCER AprIl 1, 1998 I Volume 82 I Number 7

of the smoking population). We also did notcategorizeexposure according.to different types of alcoholic bev-erages, as there is as yet no convincing evidence thatiionethanolic ingredients in alcoholic beverages areof any importance in the etiology of head and neckcancer.°4

of special interest to the Nordic countries is“Swedish” oral snuff. It is a moist, nonfermented to-bacco. mainly produced from dark Virginia tobaccomixed with Kentucky tobacco. It is a special Scandina-~‘ian product used mainly in Sweden and to a lesserextent in the’ other Nordic countries. Outside this geo-graphic area this snuff is virtually not used at all. It issometimes said that 15% of all males in Sweden useor have used Swedish oral snuff. In our material, 14%of the referents and 15% of the cases had used or werecktrrelit users of Swedish oral snuff. l’he consumptionis highest among males and the habit is most commonitt the North of Sweden. ibis geographic area is a lowincidence area for head and neck cancers. In the areas‘with tite highest incidence of cancer of the head andneck (the urban areas of Stockholm, Gothenburg. andMaltnö). tite consumption of oral snuff is lower. Incom~)arisortwith countries where oral snuff Is seldomused, Sweden has a much lower incidence of head andneck cancer, especially cancer of the buccal mucosaand gingiva.’ Within the European Union, a discussionis taking place over the role of oral snuff in the etiologyof cancer with special concern for cancer of the headand neck. ihe concern is especially important. as thistype nl oral snuff is believed to be widely used amongmale teenagers and it is often being regarded as analternative 10 Cigarelle smoking. Oral snuff containsN. ititrosamines with carcinogenic potential. Reportsfrom the U.S. have indicated an increased risk of oralcancer associated with the use of oral snuff,6-7 andtite International Agency of Research on Cancer hascoiiuludcd that nonsmoking tobacco is ha7.ardous.~‘I’hc prohibition of oral snuff by the European Unionmember nations has been urged. Sweden has so larbeen att exception to this prohibition. In our stud>’.relative risks were usually close to RR= 1. Age at start,total number of years of use, and total amount useditt a lifetime had little or no impact on RH. A highintensity of usage (>50 grams/week) was associatedwith moderately. hut not significantly, elevated risks:flIt 1.7 (95% CI 0.8—3.9) for cancer of the oralcavity and HR 1.9 (95% Cl = 0.8—3.9) for cancer oftite esophagi’s. The snuff is known to produce ulcer-ation at the place of application on the gum under theupper Tip. Also, white lesions often appear at the placeof application. There is tto clinical evidence that theselesions transform into malignancies, and the ntucoustissue nornialii.es after cessation of snuffdipping.~On

the contrary, as mentioned before, cancers of the gin-giva or buccal mucosa are very rare, with only 14 casesa year on average in the Stockholm area (1990—1993).Of these, none were located inside the upper lip. Thedifference in results between the studies by Winn etal. and ours might be related to differences in studytechniques.6-7 The studiesbyWinn et al. involved refer-ents selected from among patients admitted to hospi-tal for reasons other than cancer. These patients donot necessarily represent the true use of oral snuff inthe study base. Also, no consideration of chronic irondeficiency anemia was taken, although it is known thatchronic anemia among women can produce squa-mous cell carcinoma of the upper gastrointestinaltract. The fact thatdifferent types of oral snuff containdifferent amounts of carcinogenic agents, due to boththe ingredients used and the production process,~isalso a plausible explanation for the different results.It is noteworthy that the REt (47.5) for the use of oralsnuff found by Winnet al. was based on small numbersof individuals. Before forming public health recom-mendations or regulations, the REt for squamous cellcarcinoma of the head and neck associated with Swed-ish oral snuff compared with tobacco smoking, as wellas the possibility of other yet unknown health risks,has to be considered.

In conclusion, we confirm others’ findings of adose-dependent excess risk of cancer of the head andneck from tobacco smoking and alcohol consumptionamong Swedish males. Moderate alcohol intake hadlittleor no effect among nonsmokers. No significantlyincreased RH was found for the use of Swedish oralsnuff.

REFERENCES- Ilakulinen 1, Andersen B. Pukkata E. Schou C. Tulinius 1-I.

Trends in cancer incidence in the Nordir countries. ActaI’athol Microbio! ¡enmono! Sc.and1986:94:30—7.

2. Franceschi S. Bidohi E. Negri E. Barbone F, La Vecchia C.Alcohol and cancers of the upper aerodigestive tract in menand wonten.CancerLpiden,iolBiornarkersPier’ 1994;3:299—304.

3. Maier H, Diet.z A, Gewelke U, Heller W-D, Weidauer H. To-bacco and alcohol and the risk of head and neck cancer.Clin invest 1992;70:320—7.

4. Parrish MK, Higuchi S. Lucas IL Increased alcohol-relatedesophageal cancer mortality rates in Japanese men. ¡nl IEpiden,io! 1993:22:600—5.

5. Ranasinghe A. Macgeoch C, Dyer S. Spurr N, Johnson WN.Some oral carcinomas from Sri Lankan betel/tobacco chew-ers overexpress p53 oncoprotein but lack mutations in exons5—9. Anticancer Res 1993; 13:2065.-B.

6. Winn DM, Blot WI, Shy CM, Pickle LW, Toledo A, FraumeniJ)F. Snuff dipping and oral cancer among women in thesouthern United States. NEnglJMed 1981;304:745—9.

r

r. ~

7. Wint, 1)M, Ziegler HG, Pickle LW, GridleyG, Blot WJ. HooverltN. tit’i in the etiology of oral and pharyngeal canceramong women front tIte southern United States. CancerRes1984:44:1216-22.

8. Andetssoit(, AxéIl ‘1’. Chitrical appearance of lesions associ-ated with the use of loose and portion-bag packed swedishmoist snuff: a comparative study. Pathol Med 1988:17:1—6.

9. Mat Isson hI. Cancer registration in Sweden. Thesis, Depart.mnet,ts of oncology and cancer epidemIology, Stockholm,Sweden: Karolinska Instituict, 1984.

10. c;erhardssotl de Verdier M. Flagman U. Stetneck G. RiegerA, Nort’ll SI.. l)iet, hod) mass and colorectal cancer: a casereferent ~tud~’in Stockholm. Intl Cancer 1990:46:832—8.

11. Iterg~tront 1.. Kylbcrg E. ilagmnan U, Eriksson H-B, Bruce A.lite food conmpoSition data base KOST: the National FoodAdtttinistration’s information for nutritive values of food.V~irloda 1991:43:439-47.

12. ttre~k,wNE. t)ay NE. Statistical methods in cancer research.Volttnie 1. lyon, France: 1m,ternational Agency for Researchott Cancer, 1980.

13. J’crssomi I’-G, NoreII SL. Retrospective versus original infor-mn~,mionott cigarette sntoking: implications for epidemiologic~.iudies..4P’t ¡ l:pidcflhiOi 1989:130:705—12.

14. Blot ~Vl.Mclaughlin 1K, Winn 1)M, Austin DF. GreenbergIll. l’rectnn-Mar&in S. em at. Smoking and drinking in relationto oral and piraryngeal cancer. CancerRes 1988:48:3282—7.

IS. itundgaard 1. Wildt J. Frvdenberg M. Elbrond O, Nielsen I.(.a’.c—control smud~’of squamous cell cancer of the oral cay-tv in I)enii,ark. (~‘uJcerCauses Control 1995:6:57—67.

lb. Ilamuder’. \VI). Romhnuan XI. Interaction of tobacco and aleo-fmi ir, laryngeal cancer. Ant I Epidemiol1982:11S:371—9.

7. Ht,tltm,otn K. Keller A. The effect of ioinl exposure to alcoholut,d t(utu:ucc(J tu,, rick of cancer uf the mouth and pharynx. I(un,,, l)u 1972:25:711—6.

lH ltrrigt-rc I. (~t,~nelI’. t.eclcrc A, Rodriguez 1. Differential cf-fLU’. tul iotuaL-L-o otid alcohol in cancer of the larynx. pharynx

Rl ti un, t B Coin-er 1986:57:391 —5.hi. tu~rt’.Al. I.~t~veI. Ravmotrd 1., Berrino F, Renhan~ouE,

ltl,tu,clit-t I . t’t al. Cancer oftime tarvnx/hypopharynx. tobaccoat,d ,ilculuuh. Jut) (~‘,nc,’r1988:41:483—91.

:‘o. S1

,jn, Mil. ILr’gr’r Ii. Got-pfcrt H. F1on~WK. Newell. Gil.Squ~uruirurI’.cell r-arciumt,ut,a of tite upper acrodigestive tract.(u,,,,’, 19B8:hI:203-8.

11. lkuvIL’ I’. ~cI~,t-farl,itec;j. Zhemig 1. Maisonneuve P. Lvstifeeva

Tobacco and Alcohol In SCCHN Etiology/Lewin at al. 1375

T, Scully C. Recent advances in epidemiology of head andneck cancer. Curr Opinion Oncol 1992:4:471—7.

22. Choi SY, Kahyo H. Effect of cigarette smoking and alcoholconsumption in the aetiology of cancer of the oral cavity,pharynx and larynx. Intl Epiderniol 1991;20:878—85.

23. Merletti F, Bofetta P. Ciccone C, Mashberg A, Terracini B.Role of tobacco and alcoholic beverages in the etiology ofcancer of the oral caviry/oropharynx in Tormo, Italy. CancerRes 1989;49:4919—24.

24. Day GL Blot WJ, Mclaughlin JK, Fraumeni IF ir. Carcino-genic risks of dark vs. light liquor. Intl Cancer1994:59:319—21.

25. Kato I, Nomura AMY. Alcohol in the aetiology ofupper aero-digestive tract cancer. Oral Onco!EurJCancer1994:308:75—81.

26. Blot WJ. Alcohol and cancer. Cancer Res 1992;52(Suppl):21 19s—23s.

27. Eltwood IM, Pearson ICC, Skippen DII, Jackson SM. Alcohol,smoking, social and occupational factors in the aetiology ofcancer of the oral cavity, pharynx and larynx. Intl Cancer1984:34:603—12.

28. Schottenfeld D. The etiology and prevention of aorodigestivetract cancers. Ad,’Exp MedBiol 1992:320:1 —19.

29. Hedberg K, Vaughan TL, White E, I)avis S. Thomas I)B. Alco-holism and cancer of the larynx. Cancer CausesControl1994:5:3—8.

30. Olsen I. Sabroc S. Fasting U. Interaction of alcohol and to-bacco as risk factors of cancer of the laryngeal region. IEpidentiol conmrnuniryHealth 1985:39;165—8.

31. Olsen I. Sabroc S. lpsen I. Effect of combined alcohol andtobacco exposure on risk of cancer of the hypopharynx. ¡Epiderniol CommunityHealth 1985:39:304—7.

32. un C. un y, Wennerberg I. Akervall I. Grenthe B. MandahlN, et al. Clonal chromosome aberrations accumulate withage in upper aerodigestive tract mucosa. Mutation Res1997:374:63—72.

33. Brennan IA. Boyle lO. Koch WM, Goodnian SN. Hruban RH.Iby VI, et at. Association between cigarette smoking andmutation of the p53 gene in squamous-cell carcinonta ofthe head and neck. N Engl J Med 1995:332:712—7.

34. Tobacco habits other than smoking; Betel-Quid and areca-Nut chewing: and some related nitrosamines. Volume 37.Lyon, France; International Agency for Research on Cancer,1985.