Tobacco Documents Online

AW.ERICAN JO4.RNAL OF EPIOEWIOLOCF Vol. 121L?lo. i Cop.rtghtc. 19P+5 b)The Johns HopkuuUnnersn.v School.of'HyFiene and Public Health Pnrued in U.S.A. AIInRhts reserved PASSIVE SMOKING IN ADULTHOOD AND CANCER RISK' DALE P. SAA'DLER, RICHARD B. EVERSON AND ALLEN J. WILCOX Sandler„D. P. (National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709), R. S. Everson and A. J. Wilcox. Passive smoking in adulthood and cancer risk.Am J Epidemiof 1985;121:37-48. Overall cancer risk from adult passive smoking has been examined using smoking by spouse as the measure of exposure. Information on smoking by -spouse was obtained for: 518 cancer cases and 518 noncancer controls. Cancer cases were identified from a hospital-based tumor registry in North Carolina. Cases included all sites except basal cell cancer of the skin and were between the ages of 15 and 59 years at the time of diagnosis. Cancer risk among individuals ever married to smokers was 1.6 times that among those never married' to smokers (p < 0.01). This increased risk was not explained by confounding by individual smoking habits, demographic characteristics, or social class. Elevated risks were seen for several specific cancer sites and were not . 'timited to lung cancer or other "smoking-related" tumors. Risks from passive smoking appeared'greater among groups generally at lower cancer risk (females,, nonsmokers, and individuals younger than age 50 years), but were not limited to these groups. neoplasms; risk; smoking; tobacco smoke pollution Passive exposure to cigarette smoke has been linked~with a variety of health conse- quences in humans, including bronchitis and pneumonia in infants (1); reduced pul- monary function (2) and acute respiratory disease in children (3-5), and decreased; airway function in otherwise healthy adults (6). Several'reports have also focused atten- tion on a possible association between pas- sive exposure to cigarette smoke and lung cancer (7-10). Received; for publication January 23. 1984. and in final form April 12; ,1994. ' From the Epidemiology Branch, Biometr y and Risk Assessment Program, aational Institute of En- vironmental'Health Sciences. Research Triangle Park, NC. Reprint requests to Dr. Dale P. Sandler, Epide- miobg.• Branch. Mail Drop A3•02, National Institute of Environmental Health Sciences, P. 0. Box 12"233, Research Triangle Park„NC 27,709. The authors thank Dr. James P. Browder and the staff of the Cancer Data Base. North Carolina Me- morial Hospital for allowing the use of the tumor reFistry for case identifcation, and: David L. Shore. Karen LMilne„ and Sue W. Ward for assisting in data management and analysis. In a case-control study by Trichopoulos and colleagues (?), 51 white females with lung cancer and 163 controls from an or- thopedic service were compared with regard to smoking histories of husbands. Women who were not smokers but were married to smokers were at two- to threefold risk for lung cancer compared with nonsmoking women who were not married to smokers. An updated report of this study involving 77 nonsmoking lung cancer cases and; 225 nonsmoking controls confirmed the two- fold risk among passive smokers (8). In a prospective study from Japan; Hirayama (9) observed 245 lung cancer deaths among 91,000 married women. The lung cancer death rate for nonsmoking women married to smokers was nearly twice that for non- smoking women married to nonsmokers and was one-half that for women who themselves smokedl Emphysema was the only other cause of death to exhibit such a pattern, although the trend' was not statis- 37 , 11

38 SANDLER ET AL tically significant. Most recently, Correa et al. (10) reported a twofold risk for lung cancer among nonsmokers married to smokers. In contrast, analysis of data from an American Cancer Society study in the United States failed to demonstrate an as- sociation between passive exposure to cig- arette smoke and lung cancer risk (11). Many of the constituents of mainstream cigarette smoke which is actively inhaled by the smoker are present in reduced quan- tity in exhaled smoke (12) which is then passively inhaled by the nonsmoker. These same constituents are also contained in sidestream smoke which is released from the cigarette between active puffs and is also inhaled by the passive smoker. One might expect to find, as did Hirayama (9) for lung cancer, that for smoking-related sites, the cancer risk in individuals pas- sively exposed to cigarette smoke might fall between that for smokers and that for non- smokers. Risk from passive smoking might also be expected to be much lower than-that from direct smoking. However, some chemicals appear in higher concentration in sidestream smoke than in mainstream smoke, making the ex- posure from passive smoking qualitatively different (12, 13). The health consequences from passive smoking may, therefore, differ from those of direct smoking. These chem- icals, many of which are known carcino- gens, might lead to increased risk for cancer at sites not shown to be related to direct exposure to cigarette smoke. In comparing sidestream with mainstream smoke, Brun- nemann (in refs. 12 and 13) found 52 times as much dimethylnitrosamine, 16 times as much naphthalene, 28 times as much methylnaphthalene, 3.4 times as much benzo(a)pyrene, and 5.6 times as much tol- uene, for example, in sidestream smoke as in mainstream smoke. While the concen- trations of these chemicals are higher than in mainstreamsmoke, actual exposure from passive smoking is heavily influenced by the amount of smoke generated, the volume of ambient air, room ventilation, and the manner in which the cigarettes are smoked. Differences in the route of inhalation of sidestream and mainstream smoke might also account for differences in site-specific effects. Wynder and'Goodma4i (14), for ex- ample, have proposed that'if sidestream smoke components are inhaled through the nasal passages, gaseous components but not smoke particulates would reach the lung. As a preliminary exploration of the hy- pothesis that passive exposure to cigarette smoke may be carcinogenic, we examined adult passive exposure to cigarette smoke in relationship to cancers of'all sites. Since active and passive smokers may differ in the miz of carcinogens to which they are exposed, it is not obvious which sites might be at highest risk of cancer among passive smokers. Since active smokers are also pas- sively exposed, candidate sites might be drawn from those that have been linkec with active smoking. However, there ma} be additional sites whose relationship tc smoking has been obscured. In studies com• paring smokers with nonsmokers, passivi smokers are often included in the non smoking group. This wouldmake it diff cul- to detect small differences in risk due tr passive smoking. We report here on cance risk from passive smoking using smokinj~ histories of spouses as a measure of passivi exposure to cigarette smoke during adult hood. , METHODS Data reported here are from a study c childhood exposure to cigarette smoke an cancer risk in adulthood (29). Cases fc study were selected from the hospital-base tumor registry at the North Carolina Mea orial Hospital of the University of Nort Carolina in Chapel Hill. They included a cases diegnosed between July 1, 1979 an March 31, 1981 and assumed to be alive e of March 31, 1981. Cases were between th ages of 15 and 59 years at the time ( diagnosis andd included all cancer sites e: N O tJ G,y Qn N~ C!t ~ N

I PASSIVE SMOKING iN' ADIJLTHOOD AND ~ CANCER RISK cept- basal cell cancer of the skin. Cases were restricted to this age group to maxi- mize the likelihood of also detecting effects from childhood exposure to cigarette smoke. Individuals older than 60 years in 1981 are not likely to have ha& mothers who smoked. Patients with multiple pri- mary tumors were included only if the first primary tumor was diagnosed during the study period. Cases were mailed a questionnaire for self-completion. This mailing was followed by a second mailing and then a telephone call if needed. In addition to questions on exposure to cigarette smoke, cases were asked to identify friends or acquaintances who did not have cancer to serve as com- parison subjects. These friends were the same race, sex, and age (±5 years) as the cases. Approximately 60 per cent of the controls were identified in this manner. For cases for whom friend controls were not successfully obtained, population controls were identified by systematic telephone sampling. Beginning with the telephone numbers of the cases, the next higher or lower telephone numbers were called until individuals of the same race, sex, and age (t5 years) were found. For cases inter- viewed by telephone, the calls to identify controls were made at the same time of day.. For cases contacted by mail, telephone con- trols were chosen during randomly assigned times of day. Of 740 eligible cancer cases identified from the tumor registry, 107' (14 per cent)) die& before we could contact them. An ad- ditional 115 (1S per cent). either refused (n = 71) to participate or could not be con- tacted. In all, completed questionnaires were obtained for 518 (70 per cent) of the eligible cases. Of 518 cases, 360 (70 per cent) named friends or acquaintances who could be con- tacted' as controls. Of these, 86 per cent were successfully contacted for an overall response rate of 60 per cent. To obtain the additional 209 controls, 1,237 households were telephoned. Screening data (age, race, 39 sex, and cancer history of household mem- bers) were obtained for 988 households (80 per cent); 224 (23 per cent) of these house- holds had a qualifying member. Fifteen (7 per cent) qualifying telephone controls re- fused to participate. The overall: response rate for selection of telephone controls was 75 per cent (80 per cent x 93 per cent). Although not shown here, d'ata were ana- lyzed separately by control selection group, and;the adjusted results were identical to those obtained when the control groups were combined. Procedurally, the control selection pro- cess involved one-to-one matching. This was done to allow the selection of popula- tion controls without having an enumer- ated sampling frame. The analyses pre- sented here are for unmatched data to max- imize the study sample size following losses due to missing data on exposure. In most comparisons„ the factors used in control selection are taken into account by adjust- ment procedures. Parallel analyses for matched pairs were carried out. Although not presented here, the results were sunilar.. Passive exposure to cigarette smoke dur- ing adulthood was estimate& from a ques- tionnaire report of the number of years of marriage during which a spouse smoked. Subjects were considered exposed if they had a spouse who smoked regularly at any time during their marriage. Regular smok- ing was defined as smoking at least one cigarette per day for as lbng as six months. The nonexpose&group consisted of persons married to nonsmokers and persons who never married. The quantity smoked wass reported as the average number of ciga- rettes smoked per day by' a spouse while married to a study subject. For the analysis of questionnaire data, odds ratios were calculated{ and the chi- square test was used' to assess statistical significance. Combined estimates of the odds ratio (OR) in stratified analyses were obtained using the Mantel-H'aenszet tech- nique (15). The method of Gart (16) was used to obtain 95 per cent confidence limits

I 40. tiANDI.F.R F.T Al- for the combined estimates of the odds ratio. When the 95 per cent confidence •lfmits did not include unity, the odds ratio was considered statistically significant (p < 0.05). Level of education was reported as number of years of school completed and occupation was given as usual occupation. For stratified and adjusted analyses, age and level of education were treated as cat- egorical variables with four levels of age (<30, 30-39, 40-49, and 50+ years) and three levels of education (<12 years, 12 years, and >12 years). RESULTS The distribution of cases by cancer site is shown in table 1. The young age of cases, the referral nature of the hospital, and the fact that the study was limited to living cases account for the distribution of cancers seen. There was a predominance of breast cancers, female genital tract cancers, and leukemia and lymphoma, and a relative lack of respiratory tumors. Eligible cases with respiratory cancer were significantly more likely to die before they could be contacted for this study. Cases and controls are compared in table 2. Cases and controls are, by design, dis- trihuted similar?y by race and sex, with 70 per cent of each group white and 67 per cent female. The mean age of cases (43.6 years) and controls (43.5 years) is also sim- ilar. Level of education differs sigaificantly betweeti groups; 45 per cent of cases and 36 per cent of controls never graduated fromm high school. Level of education, therefore, is taken into consideration in most anal- yses. On the other hand'; cases and controls do not differ by broad occupational cate- gory. A greater proportion of controls never married (16 per cent vs. 13 per cent), but this difference is not statistically signifi- cant. Cancer cases and controls also do not differ in reported smoking histories. In part this is due to the relative absence of lung cancer cases and to the method of control selection. Sixty per cent of controls are friends of cases, and the smoking habits of individuals who are friends may be similar. When only cases with population controls are included, 57 per cent of cases and 46 per cent of population controls were smok- era. The overall crude cancer risk for individ•, uals ever'inarried`to siriokers'is times that for those not'isiarried to smokers (p <' 0.01Y (table 3).' Adjusting independently, TASi.E I Distribution of cancer cosus bv aite o/ primary tumor and studi• status Includtd Refused'or lost Site 1CDP No. No. 1`i) No. tT.) Lip. ord cavity; and pharynx 140-149 22 (4) 15 (7) Digestive organs and peritoneum 150-159 41 (8) 16 (7) Respiratory and intrathoracic organs 160-165 32 (6) 43 (19)' Bone, connective tissue, and skin 170-173 42 (8) 13 (6) Breast 174, 60 (12) 16 (7) Female genital organs . 179-184 175 (34) 61 (28l. Prostate 185 10 ('_') 0 (0) Testia 186 6 (1)' 3 11) Urinary tract 188.199 6 (1) 9 (4) Eye. brain, and other nervoue s,ratem 190-192 :S8 1ll) 20 191 Thyroid' and other endocrine gland! 19'1. 194 : L' l51 3 (1) Lymphatic and hematopoietic tissue :Z00-207 52' (10) 13' (6) Site unspecified 199 7 (1); 10 (5): All sites SIR (100.0) (10M01. •p<0:01. t ICD, lntrrnaliond Cf6si(icotion of Dr..ea.es. Ninth Revision.

PAS51VE SMOKING IN ADULTHOOD AND CANC£R' R1SK 41 TABLE 2 also appears limited to individuals who are ContAariaon of casts and controls younger than age 50 years. Factor Cases Contrvls r-O. I•rl No. I'7) Total 518'(100) Sd811U01 Age <30 96(19) 99(19) 30-39 89 (1 -,) 105 (20) 40-49 13'_ 125) 121 tt?3) 50+ 201 (39) 193(37) Race Nonwhite 153 (301 153 (30): White 365(i0); 365 (i0) Se: Male 169(33) 169(33) Female 349 (6T) 349 (6.7) Maritalistatus' Never married 6503), T9(18) Ever mamed 444 ' (9-1) 410(94) Eduntion• <12 years 233(45) 186136) 12 years 137 (27) 186(36) >12 years 147 (28) 146 (28) Occupationt Blue collar 1"^_ (351 194 (38). White collar 192 (39) 175(34) Housewife 1181241 131 t_61 Unemployed 8(2) 1142) Smoking Never 235 1451 247 (48) Ever 283 (551 2"1 (S2) Current 154 130) 166 l3_): Past 129 (rl 105(20) , 'Nine cases and 29 eontrols did not report merital status. t One case did not report years of education. = TwentyeiRht cases and seven controls did not report occupation. and'in combination for sex, age, race, smok- ing, parental smoking, education, and oc- cupation does not change this finding. Can- cer risk from passive exposure to cigarette smoke appears greatest for females and for individuals who are not themselves smok- ers, with statistically significant risks lim- ited to these subgroups. There are no ap- parent subgroup, differences in risk with race or occupational category (blue collar or white collar), although risk appears greater for individuals with at least a high school education. Cancer risk in relation- ship to passive exposure to cigarette smoke Cancer risks from passive smoking among smokers and nonsmokers are shown separaroely in table 4. Risk is clearly ele- vated among nonsmokers, with the twofold risk significant after adjustment for age, race, or sex. Risk is also elevated among smokers, but the 30 per cent increase in risk is only of borderline statistical signifi- cance. Among nonsmokers, risk does not differ with race, but the risk from passive exposure is statistically significant only among females and among individuals be- tween the ages of 30 and 49 years, although it is also elevated for males. For smokers, risk is significantly elevated among females and whites. The twofold risk related' to passive exposure among individuals younger than 30 years (table 3) is due to risk among individuals who are themselves smokers (OR = 2.3): (table 4). The lower risk among nonsmokers in this age group (OR = 1.4) contrasts with the greater risk among nonsmokers in the other age cate- gories under age 50 years. This suggests that the cumulative exposure through pas- sive means, albne, for this young group may be below that which would pose a risk. For most cancer sites, the number of cases is too small for meaningful' site-spe- cific analysis. However„statistically signif- icant risks in relationship to passive smok- ing are seen for breast cancer, cervical cancer, and endocrine cancers. Odds ratios adjusted for possible differences in the dis- tributions of age and level of edilcation are shown in table 5 for smokers and' non- smokers combined. The twofold risk of breast cancer shown in table 5 is not substantiaTly changed by adjustment for education, race, age, smok- ing status, or parental smoking. Breast can- cer risk is greater among younger women (OR = 3.4 for women G50 years vs. OR - 1.1 for women _50 years) and those with at least a high school education (OR = 3.3 vs. OR = 1.0). The twofold risk for cervical cancer

I 42 SA.NDLER • :L TAaLF S Ooernll cancer risk,(rovn smoking bv spouse, ttdJusted srporotelyt for poterstic!'confoundind fartors % s:Do..di 95Se CLt oe adju d F.ctor CAMS5 (n - 509) ~~`s (n - 4891 Cnd! ORt Adjwced'OR au OR ~ Crude risk 5S 43 1.6" (1.3, 2.1) Sex Male 35 26 1.Y` Female 65 51 l:n" 1.7" (1.3.2.2) Age <30 1 39 24 2.0• 30-39 55 39 1.9' 1.6" (1.3, 21) 40-49 67 47 2 1" 50+ 56 54, 1.2 Race Nonwhite 50 34 1.6' Khite 58 45 1.7" 1.6" (1'.3,21) Smoking Nonsmokers 52 34 2.1" Smokers 58 51 1-3 1.6" (13 211 Education <12 years 55 52 1_1 12 years 61 35 2.9" 1.6" ( L.2, 2I) >12 years 50 41 1.5 OccupationI Blue collar 53 40 1.7 White coller 53 40 L7* l. i• (1.2, 2-n ~ Either parent smoked No 55 41 1.8'• 1.8*' (1.3, 2.3). Yes 5-1 44 1.7•' •p<0.05. ••p<0.01. • OEt: odds ratio, after multiple adjustment for age, race. sex, education, and smoking status - L7, 95% confidence limits (CL) - (1.3. 2.3). = Spouse ever smoked while marriedto study subject. ~ Numbers of cases are not equivalent to those shown in table 1 because of missing values for education or spouse smoking. I Excludes housewives, those unemployed, and those missing data on occupation. among individuals passively exposed to cig- risk of endocrine tumors among exposedo arette smoke is also not affected by adjust- individuals, which remains after adjust- ment for age, race, smoking status, or,? ment for potential confounding variables. smoking by parents. The estimated risk is In subgroup analyses, risk is significant for reduced somewhat by adjustment for level younger individuals, nonsmokers, individ- of education, but there is no clear pattern uals with a high school education, and in- of risk with level of education. As for the dividuals whose parents did not smoke.. other sites, risk appears greatest among Although. the_ number of lung cancer younger women (OR - 2.9 for women <50 cases is smali"ei€''="=22~~uag'''c'snctS risk years vs. OR - 0.9 for women >50 years). from pase exposure to cigaFe,tAsmgking Odds ratios are statistically significant for is examined in table 6 because of current whites and for nonsmokers, but the mag- interest in this site. The overall crude risk nitude of the risks for nonwhites and for of 1.9 is not statistically significant. ow- smokers is similar. ever, thee odds ratio ~> ~~ There is also a statistically significant a and for nsm ~ers"T : x, ~ N. O 2V GW W ~ N UZ 0?

PASSivE SMOKING IN ADL'LTHOOD AND CANCER R1SK 43 TAat.E 4 OoernfU cancer risk /rom possiue exposure to eigarrtte smoke among smokers and nonsmokers, adjusted for patentia/ confounding factors Nonsmoken.ln - 1G6) Smok'ersfn.- 832). Factor Casest Controls Cnide 0~ Ad*ted OR2 (s5`: Casesr Controls Cnde OR AdjuvW ORI (95% No: (11 ex9osed) lyo_ ir; etpo"ed) CLSI No: ( : eiPosed), No. t!" etPosedl i CLS) Crude risk 231 l52) 235(34) ' 21" 2' 8 (58) 254(51) 1.3 (Q.4, 3.0) (0.9, 1.9) Ase <30 45 (22) 5807) 1.4 50(54) 38(34) 2.3 30-39_, _ 36 156) 49(31) 2:8' 2_0•" 51 (55) 48/48), 1.3 1.4 40-49 63(681 48(33) 4.3" (1.4, 29). 67(64) 66(58) 13 (1.0, 19) 50+ 87 (53) 80(49) 1.2 110 (58): 102(55) 1.1 Sex Male 39(13) 57(9), LS 2.0*' 128 (41) 102(36) 1.2 1.5' Female 192(59) 178 (42)1 2.0" (1.3.2.9) 150(73) , 152(61) 1.7' (1.0:2.1)' Race Nonwhite 72(53) 8a (31)' 2.5'" 2.0" :~ (47) 63448) 1.0 1.4 White 159 (510 152(36) 1_9" (1A. 3.0) 201(631 191'152) 1.5' (1.0.1.9) . • p < 0:05. "• p < 0.01. t Numbers of cases are not equivalent to those shown in table 1 because of missing values fon spouse smoking. j OR, odds ratio. S CL, confidence limits. cally significant even though they:are based on small numbers. Since only two lung cancer cases were nonsmokers and only seven were females, it is not possible to examine lung cancer in greater detail- Risk is also elevated among younger individuals and among those with at least a high school education. Among nonsmokers alone, risks were sig- nificantly elevated for endocrine and cer- vical cancer, despite the loss of power from reducing the already small number of cases for site-specifc analysis (table 7). The two- fold risk for breast cancer seen overall is also seen for nonsmokers but is not quite statistically significant. Among smokers, the odds ratio is statistically significant for breast cancer only. However, the approxi- mately twofold risk for cervical cancer is similar to that among nonsmokers. There is no clear dose-response relation- ship for all cancer sites combined or for specific sites in relationship to either the number of years married to a smoker (ad- justed for age) or the average number of cigarettes smoked'per day: This is true for smokers and for nonsmokers and also when analysis is confined to those younger than age 50 years, to whom an effect of passive smoking appears limited. DISCUSSION We have found a significantly elevated overall cancer risk for individuals passively ezposed'to cigarette smoke.. This cannot be readily explained by a number of other factors, including individual smoking hab- its and two measures of social class: edu- cation and broad occupational category. El- evated risks were seen for several specific cancer sites and~ were not limited to lungg cancer or other "smoking-related" tumors. These findings might relate to other factors we have not measured or to deficiencies in study design. However, we have not been able to identify a possible confounder or a bias of selection or recall that could have caused the difference in smoking patterns of spouses between cases and controls. Study subjects and interviewers were told

44 SANDLER ET AL. T/tst:E 5 Cancer ruk from pauiue ezposure to cigarette smoke, adjusted for age and education, oll sits combined and specif,irsites Site CeYf cnde ORS Ad' m.d OR; 95% CLI on adPusted Nat lfie esyo..di; p ORt All sites 508 (55) 1.6" 1.6" (1.2, 21); Lip. oral ca+ity, and pharyna 22155) 1.6 1:1 10.4; 3.0) Digestive system 39(51) 1.4 1.0 (0.5, U) Respintorysystem 32(50) 1.3 1.0 (0.5.2.4) Lung 22(39) 1.9 1.5 (0.6, 4.3) Bone, connective tissue. and skin 42(36) 0.7 0:7 -- (0:3,1.5) Brrsstl 59(69) 2.2" 1.B• (1.0.3.7) . Female genitai'avstem4 170(661 1.9" 1.8" (1S. 2.8) Cen•ix 10I(67) 2:0'• 1.8' (1.1.3.2) Prostatel 10 (30i 1.2 0.8 (0.1. 3.9) Testisl 5440) 1.9 ~ 2.6 (0.2, 49:9) Urinary tract 6(50) 1.3 Ll (0.2. 7.6) Eye. brain, and other ner vous systam 38132) 0.6 0.7 !0 3. 1S) Endocrine 26/65) 2.5' 3.2" (1.4, 9.4/ hlematopoietic 52144); 1.1 1.3 (0.7.2:5) Other 7(57)' I.8 1.8 (0:3, 10.4) •p<0:05. •• p < 0.01. t Numbers of cases are not equivalent to those shown in table 1 because of missing values fot education or spouse smoking. ; For comparison. 210 of 489 controls (43 c) were exposed. ; OR. odds ratio. I CL confidence limits. t Sex•specific comparison. Uf 330 female controls. 51% were exposed. Of 159 male controls. 26 e.rere ezDased. simply that the study was designed to look at smoking patterns in families. Cases and-controls were similar with re- gard to their own smoking histories. This was partly because of the choice of friend controls who tended to have similar smok- ing histories and because known smoking- related sites were underrepresented' in the case population. Cases included in the study were generally younger than those with smoking-related tumors. Unavoidable delays between case identification and completion of interviews also contributed to the lack of smoking-related'cancers. Per- sons with lung cancer and other smoking- related' tumors were more l'ikely to have died before they could be interviewed. In addition, because of the special interests of physicians at the hospital from which cases were identified, breast cancers and gyne- cologic cancers were overrepresented. As a result of this unintentional matching on smoking status, risks from passive smoking and direct smoking cannot be compared. The route of exposure for the passive smoker is via inhalation. Reports of effects on upper respiratory tract function (2-6) are consistent with this. There ha3 also been a report of mutagens measured: in the urine of passive smokers (12), indicating that components of cigarette smoke enter the bloodstream and are circulated throughout the body of the passive smoker. Another report indicated that enzyme ac- tivity can be induced by passive exposure to cigarette smoke (18). These findings are tentative, but do suggest that an overall increase in cancer risk or an increase in risk for specific nonrespiratory sites follow- ing passive exposure to cigarette smoke is plausible.

PASSIVE SMOKING IN ADIDLZ'HOOD AND CANCER R15K 45 TABLE 6 (,ung cancer risk from passive exposure to cisarettr smoks, ad1usted'(or potmtiol'con/oundin; factors Fa Cav+ Cntdk OR+ Adjusted ORt 955 C1.S an adjuwed ctor No. i re nPosedt: OR* Crude risk 22 (59)~ 1.9 (0.8, 5.0) I - Sex Male 15 1101 1.9 Femsde 7 (100)' a" 3.4' Smoking Nonsmokers 211001 Q' Smokers 204551 1.2 1L5 (0.6, 3.91 AV -- <50 5(80) 6.7 50+ 17 (53) 1.0 1.5 (0.6. 3:8) Education <1_'sYan 15 (47): 0:8. 12 yean 4 1100) a" 1.6 (0:6, 4.4) >12 years 3(67) 2.8 ~ Either parent smokedl No 6150) 1.5 1-5 (0.5.,4.8) Yes 9(56) 1.6 •p<0.05. ••p<0.01. • OR odds ratio, j CL confidence limits. ~Numben reduced because of missing data on parentrlsmoking. TABLE i i Canctrri.k from pa....itv rzposurr to ci,pontte smoke among smok.rx and nonsmokers: selected sites Nonsmokers Jmokers Site ivot OdBs ratio t95 : CL+t h~ ~ Odds ratio t95~ CL+1 Lung 2 zt 20, 1.2 (0.5.2:9)' Breaat 32 2.0 (0.9. 4.3)' 27 2:8' (1.0. -1:6) Cervix 56 2.11 11.2. 3:91 45 2_0 (0.9.4.11 Endocrine gi5nds 13 4:4' (1.2. 17.4) 13 ll5 (0:4.,5,5) 'p<0.05. • CLL confidence limiu., tp- 0.051. Our study was intended to consider a range of effects similar to what might be measured in a prospective study of a cohort of individuals who are passively exposed. Such an approach serves to single out sites which appear to be important as well as to investigate whether passive exposure might increase susceptibility to additional insults, thereby increasing cancer risk overall: Two reports in the literature use prospectivelyy collected data (9, 11). One of these (11),, however, does not provide data on cancer risks at sites other than the lung, and nei- ther report provides data on overall cancer risk from passive exposure to cigarette smoke. Data from the Japanese study re- cent3y presented by Hirayama, however, indicate that cancer risk may be increased at sites other than the lung and that risk may not be limited to smoking-related sites. (Hirayama, personali communication: pre- sented at Hawaii Cancer Conference, 1984):. For this study, passive exposure during adult life is determined from a question-

46 SANDLSR ET Ai.- naire report of the number of years of mar- ried life during which a'spouse smoked, Misclassification of exposure status is likely for individuals who never marriedbut have lived with other persons who smoked. Slightly more controls than cases reported never marrying, which might lead-to differ- ential misclassification. However, we rean- al'yzed our data, excluding subjects who never married, and found the results to be the same. When only married subjects were included; the odds ratio for cancers of all sites combined was also 1.6. We made no allowances for multiple spouses, other members of the household who smoke, or passive exposures which occur outside of the home. Quantity smoked, too, is an ap- proximate measure. The reported number of cigarettes smoked per day by the spouse is simply the average daily amount smoked during that time period: No allowance was made for changes in smoking habits of the spouse over time or for time since last exposure if the spouse did not smoke during the entire married interval: Nonetheless, we found smoking by spouse to be significantly associated with overall cancer risk. The odds ratio of 1.6 was not substantially altered by adjustment for age, race, sex, smoking status, educa- tion, or occupation. Risk was limited' to individuals younger than age 50 years, who were at approximately twofold risk. Risk was also greatest for females and non- smokers, although not entirely limited to these groups. When smokers and nonsmokers are con- sidered separately, the twofold risk among nonsmokers is highlysignificant and is not altered by adjustment for potential con- founding factors. The 30 per cent increase in risk among smokers whose spouses also smoke is only of borderline statistical sig- nificance, but is also unchanged by adjust- ment for other factors. The groups for whom risk from passive smoking appeared greatest are those groups generally at lower cancer risk overall. It may be that the smalll risk imposed by passive exposure during adult life is difficult to detect statistically in individuals at risk from other causes. Also, women who smoke may tend to smoke less, etart- later, and inhale differently than men. This would allow for a greates impact -of passive exposure among women; regard- less of their own smoking status. In addi- tion, very few nonsmoking men are married to smokers, making it more difficult to de- tect a risk among males. In our data, only 10 per cent (10/96) of nonsmoking males were married to smokers, whereas 51 per cent (189/370) of nonsmoking females were married to smokers. The increased cancer risk from passive exposure was not limited to sites generally thought to be smoking-related (12, 13). In fact, because of our case selection proce- dures and delays in interviewing cancer cases, individuals with cancers of smoking- related sites were only a small proportion of total cases. If cancers of the esophagus, respiratory tract, oral cavity and pharynx, urinary tract, and pancreas are designated smoking-related, the odds ratio for amok- ing-related tumors is 1,3, whereas the odds ratio for other sites is 1.7 (p < 0.01). Evi- dence is accumulating that cancer of the cervix should also be included among those sites that are smoking-related (19-21). When the cervix is included, the odds ratio for smoking-related sites is 2.0 and for other sites is 1.5, both of which are statis- tically significant. ' Only 22 lung cancer cases are included in this report, with an odds ratio of 1.9 among passive smokers. Although not sta- tistically significant, it is consistent with the level of risk reported in other studies. For women and for nonsmokers, the risk of lung cancer among those passively exposed was significantly increased despite very small numbers. The odds ratio for individ- uals under age 50 years was of borderline significance. Hirayama (9) reported a two- fold risk for women married to smokers and found that risks were also greatest among younger women (as measured by husband'i age). While Garfinkel (11) didn't find ar' O r>,~ c.~ ca rJ~ ~ ~ O