Tobacco Documents Online

Int. J: Cancer:,39, 162-169 (1987), •© 1987 Alan R. Liss, Inc: MEASUREMENTS OF PASSIVE SMOKING AND ESTLMATFS OF LUNG CANCER RISK AMONG NON-SMOKING CHINESE FEMALES Linda C. Koo', John H-C. Ho'-, DaisySAw3 and'ng~-yee Hot t Dept: of Conmtuniny Medicine. University of Hong Kong, Hong Kong; ZRarliothertapy Dept., Baptisr Hospital, Kowloon, Hong Kong; and 3Listitute of Patkology,, Queen Elizabeth Hospital, Wylie Road. Kowloon. Hong Kong. Lifetime exposures to environmental tobacco smoke from 28!000 inhabitants/kmi with only 8 m2 of avatlable living the home or workplace for 88 "never-smoked^ female lung cancer patients and 137 "never-smoked" district controts were estimated in Hong Kong to assess the possible causal reladon- ship of passive smoking to lung, cancer risk. Relative risks based on the husband's smoking habits, or lifetime estimates of total'years,,,tot,al hourr; mean hours/day, or totai cigarettes/ day smoked by each household smoker did not show dost response results: Similarly, when such categories as mean hoursLday, or earlier age of initial exposure, were combined with years of'exposure, there were no apparent increases in relative risk. However, when the data were segregated by histologint type and location of the primary tumor, it was seen that peripheral tumors in the middle or lower lobes, or„ less strongly,, squamous or small-cell tumors in the middle or lower lobes, had increasing relative risks that might indicate someassociation with passive smoking exposure., Epidemiolbgicall data linking passive smoking with lung cancer among non-smokers have been controversial. Six stud- ies (Hirayama, 1981, Trichopoulos eral:, 1981;,Correa et a1:,,, 1983;,Knoth etal., 1983;,Miller, 1984; Garfinkel eral:, 1985) found significantly elevated relative risks (RR') inithe range of 2.0 to 3'.5 based' on the smoking habits of the spouse. Five other studies (Garfinkel, 1981; Kabat andWynder; 1984;,Chan and' Fung, 1982; Koo et al:, 1984; Wu en al:, 1985) two of which were condut:ted',in Hong,Kong; didinot find significantly elevated RR from inhalation of sidestream tobacco smoke. Four of these epidemiologicall studies (Hirayama, 19811; Tri- chopoulos er al.,, 1981; Garfinkel, 1981i;, Chan and', Fung,, 1982) defined exposure solely by two questions: whether the spouse smoked (yes/no), and, the number oflcigarettes smoked per day by the spouse. Five other studies (Correa et af., 1983; Miller, 1984; Garfinkel ct'al., 1985; Kabat and Wynder, 1'984; Wu et al., 1985) also included questions about whether invol- untary smoke exposure had occurred at work (y,esJno); andlor whether the parents has smoked' (yeslno). Such data seem rather crude indices of exposure, providing only very indirect information on the degree and amount of exposure. Furrther- more, although, spouse(s),, parents, or co*wonkers might havee smoked, the actual degree of contact of the non-smoker with these smokers could have been very low, or even nil (Fried- man er al., 1983). In our detailed studies (Koo et al., 1983, 1984) of passive smoking exposures, smoking parents or spouses were sometimes recalled as inflicting, little or no ex- posure on the subject. In those cases where, for example, the husband smoked,but lived, separated fromi the wife„then our study counted such wives as unexposed, subjects. Among our never-smoked subjects, this was found to be true for 3' cases and 3 controls., In order to assess the possible causal relationship of passive smoking, to lung cancer risk, data from detailed life-historyy exposures that were elicited ini intensive 1_5- to, 2-hr tape- recorded interviews of never-smoked female lung cancer cases and district controls have been ianalyzed. Emphasis is placed on the consistency of the data, the strengths of the R:12„ and whether dose-response relationships were presenti. This study of the effects of passive,smoking,is particularly pertinent to Hong Kong because it is one of the mosticrowded' urbani environments in the world. Its urban density averages space per person. MATERIAL AND METHODS From 1981-3, 88 never-smoked'female lung cancer patientss and 137 never-smoked female district controls were inteo- viewed as part of a larger retrospective study of female lungg cancer in Hong Kong covering 200 cases and 200 controls. In the original,'study, patients were matched with an equal number of healthy controls by agest5' years), district of residence (1'1=3i3); and housing type (public or private housing), the latter being,an indication of socio-economic status. Details of subject selection, lung,cancer histological typing, and method of conducting the interviews have been discussed elsewhere (Koo et al:, 1983, 198i1). Never-smoked subjects were def ned as those who had smoked less than 20' cigarettes in the past. All data on passive smoking exposures were double-checked with other data elicited in the life-history interviews, espe- cially residentiali patterns since birth (i:e; where they lived, type of housing, number of rooms, number of co-habitants, etc.), occupations, and marital'life to reduce errors in estimat- ing exposure levels. Among the never-smoked subjects, the mean age of the patients was 57:81(sD 10.81) and!t;hat, for the controls was 59.3 (SD 9.94). This sample included 60 who were widows and 3' who had never married; none had married more than once. In the desigp of the interviews, separate data were collected to take into account that within the life-histories of the sub- jects, sidestreami tobacco smoke could originate from: (a) different people who smokedi in the presence of the subject; (b) different places frequented by the subject; and'{c) different types of' tobacco. Persons who smoked included related andd unrelated' members of the household, and even co-habitants who, shared an apartment unit (if their tobacco smoke was noticed by the subject). It was difftcult, to qpantify exposure levels from places that could have varying daily amounts of environmental, tobacco smoke and were occasionally visited by the subject,such as cinemas, while playing majong, or in transport vehicles. This analysis will only take into account exposures that remained relatively regular during the lifetimes of the subjects i.e. from exposures at home and the work- place(s). Among our subjects, tobacco smoke mostly origi- nated' from cigarettes smoked by household' members, and, from pipes (water and regular) smoked1by parents or in-laws. In addition to data based on the husband's smoking habits, 4 other measurements of passive smoking were evaiuated: (a) total years of exposure, (b) total' hours of exposure, (c) mean hours/day of exposure,,and (d) total cigarettes per day smoked by each household member weighed by their years of'expo- sure. These measures should'be a more accurate reflection of past lifetime exposures than simple:questions based on,whether the spouse or parents smoked (yes/no), or whether environ- mental tobacco smoke was encountered in the workplace (yesA no). Received: June 24, 1986 and in revised forrnScptember 19, 1986.

The total years of exposure were derived from adding the years during,which tobacco exposure occurred in the home or workplace. Exposures of 6' or more months were rounded,off to the next,year. In the home environment, household smokers were only counted if the subject recalled that they had smoked in her presence. Where exposure was concurrent, as in the case of, both parents smoking, or exposure occurring at the home and workplace, then the years were not added: The total hours of'exposure were calculated by multiplying the average hours/day of exposure by the years of exposure from each household smoker4 or the amount of exposure at eaclih workplace. Each of' these sources of exposure was then added together for each subject. The hours were nou added' for exposure to simultaneous smokers: For example, a husband and, son smoking at the -same time for 1 hr would only be counted as I; hr. The mean hours/day of exposure were derived by adding the hours/day of home and workplace exposures and dividing, this figure by the age of the subject. This figure approutimatess the average number of'hours of exposure per day experiencedl by the subjFct, spread'over her lifetime. A weighted average of the totali cigarettes per day smoked by each household member was calculated from the summa- tion of the usual number of cigarettes smoked throughout thee day by each household member multiplied by the years that each,lived with the subjFct, divided by the total'years during which eigarette exposure had occurred in the home. This figure may give a better indication of'the intensity, of cigarette exposure in the home than one simply basedlon the numberof cigarettes smoked per day by the husband, because it accounts for otherhouseholdlsmokers and the years that the subject'was exposed to eachi smoker. This figure excluded exposure from pipe smoking and the cigarette consumption levels of ctr workers because of difficulties in quantifying those amounts. Of the 88' patients, 83 were typed histologically. Among the remaining S cases; biopsy or cytologic materials revealed that malignant cells were present, but they were too undifferentia- ted or unspecified for categorization by cell'ty,pe: Chest radio-- graphs were examined for all cases, andl the site of the primaryy lung tumor was classified by its location inithe bronchial tree, andl whether it was centrally or peripherally situated. Ih this analysis, the lingula, was classified as eqpivalent to the middle lobe, and peripheral tumors were defined as those located beyond the segmental bronchus. Statistical analyses included the calculation of t?,Ft , as the crude or adjusted odds ratio andi tests for trend (Breslow and Day, 1980). Adjusted odds ratios were estimated by the use of a conditional logistic regression package, PECA~f„ (Lubin, 198II)1which was based!on NcM matching,by strata defined by district (N=34) and housing type (public or private),. To take into account: the effects of potential confounders which af- fected the RR'estimates, adjustments were made for age (<50, 50169; 70+), any formal schooling (yes/no), number of live births, andlyears since exposure to cigarette smoke.had ceased in the home or workplace: The exact values were used for the last, two variables. Because the resulting large numbers of matching,strata in the adjusted odds ratios may lead'to unstable results, both crude and adjusted ItR were,presented for all risk analyses. The Mantel'-Haenszel test for trend was performed on all the crude odds ratios using,the midpoint,of each interval„ whereas the trend test of the logistic parameters was based on each variable as a continuous exposure factor. (Table I): In response,to the question of whether the husband' had smoked cigarettes im the presence of the wife, the crude and adjusted RR' were both a non-significant 1.6. RR for the usual number of cigarettes smoked per day by the husband did not indicate increasing risk with higher smoking levels, and the trenditests for the crude (p=0.10) and adjusted (p=0:43) RR were not signiftcant.. Likewise, when the data were analyzed in terms of'cigarette smoke exposure during childhood/adulthood, or by the num+ ber of smoking co-habitants, as in the study of Sandler er al: (1985) (Table 11), no consistent pattern emerged. RR at the higher levels of exposure, i:e., both childhood and adulthood, or 2+ smoking, cahabitants; were found to be lower than those at lower levels of exposure. Lifetime e,rrposure measurements When the crude and adjusted odds ratios were calculated for the 4 lifetime exposure measurements, the RR for the inter- mediate exposure kvels of mean hours/d'ay ( L.94 and' 4.10), and cigarettes/day (1.57 and 2.56) were signifucant (Table M). However, with the exception of total-years, all of the RR (0.9- 1_4) at the high exposures werecbelow those of low or inter- mediate levels. Even for total, years, the Mantel-Haenszel linear trend test (p=0.55) for the crude RR, and the trend test for the logistic adjustedi parameters (p=0.23) indicated that the pattern was insignificant. When the crude and adjusted RR are compared (Fig. f); the adjustedl RR for these measurements showed RR fluctuating betweeni wider ranges ofi 1.0 to 4,1, yet both lacked evidence of a consistent dose-response pattern. Intensity As a measure of'intensity, RR'were calculated to see whether there was a direct relationship between increasing years and' mean hours/day of exposure in a 2'x2 table (Table:IV)_ Start- ing with the top left-hand square which was the group with~thee lowest exposure levels; one would expect RR' to be higher in all the other squares, especially the one at, the lower right, because it had the highest years and mean hours/day, of expo- sure. However,, the crude R<If at this highest intensity level was, only 1.07, and the category with, the lowest intensity values (top left) had the highest adjusted RR of any of the other groups. A similar pattern emerged if total hours or cigarettes/day were substituted for mean hours in this analysis.. Age of initial exposure We had previously found! no difference in the age at which, passive exposure had started (Koo er af., 1'984'), To see whether earlier age of initial exposure combined with higher years of' exposure were related with increasing risk, IfR were calcu- lated!for cigarette exposures in,aZx2 table (Table V). Again, we did not'see any pattern suggesting a dose-response, relt+tion- ship. The top left square with the least years of exposure and older age aoinitial exposurcha&the highest crude and adjusted RR. Similar results were obtained! if', the years and age of exposure included all types of environmental tobacco smoke; i.e. from cigarettes and pipe.. Histological type The cases were divided: into two groups, those with squa- mous or small-cell lung tumors, and those with, adenrncarci- noma or large-cell lung tumors: This division, was made because the former group was previously found in Hong,Kong to be more related to active smoking than the latter (Koo ctt al: 1985). Five cases with mixed cell types and 5 with unspec- xESULTS ified cell types were excluded from the analysis. Tb allow comparison of the results of' this study with those Although i none, of the crude or adjusted RR or trends by done elsewhere, exposures based omthe husbandrs cigarette histology were found to besignificant„ircan, be observed that. smoking habits were analyzed for the ever-married women a dose-response patterm seemed to be more apparent among

TABLE I - HUSBAND'S CIGARETIfES')NOKING H/1BITS.AND RR FOR LUNG CANCER AMONG EVER-MARRIED WOMEN Expoauee Husband ever smoked?2 (Ifo. Number.of eard.w number of mntmis 35/70. Yes 51 /66. Cigarettes/day smoked by husband 0 32/67 1.00 1.00 1-10 17/15 2.37'(1.03„5'.9l) 2.33'(0:912; 5.92) 11-20 25135' 1.50 (0.87„3:64) 1.74 (0.81, 3.75). 21 + 12119 1.32 (0.45„ 2:63) 1.19 (0.46; 3.03). tAdjusrcd'for age. numberof'live binhs: schooling (+f-), and yprs since e:posure to cigareae smwke oessed in rbe some or wrkpla«. =HusbaM Wmoked iin t)ie,prcsenee of'tAe rife. 3 icases and 3 conerols were oot exposett ito the ei+pumrs ofd+eir: husband. Exposure TABLE'tI - RR FOR LUNG'CANCER FROM HOUSEHOLD EXPOSURE TO CIGA'RETIIE SMOKE ?)unmber o(eases! /amAberiof eofKfUlt Cnde RR (95% ~CI)-~~ 1.00 Crude~ RRl (95!l. ~,.CI): Adjusrod RRt'(95 SG~ Cl) 1.55 (0:94'', 3.08) 1.64 (0:87, 3':09) No exposure 27/49' 1.00 iL001 Childtren only? 2/3 1.21 ('-) 2.07 (0.5i1; 95.17) Adulthood only4' 57f77 1.34 (0:84, 3'.0I) 1.68 (0:62; 5.45) Both childhood + adulthood 2Y8 0.45 (0_,111L 3.32) -- 0.64 (0:57; 5,.85) By number of smoking co-tiabirantss' None 27149 1.00 1.00 11 48/68' 1.28'(0:82; 3.25) L.73'(0.57, 6.35)'. 2+ 13120' 1.18'(0.57; 3.65) 1.35 (0:64, 5:03) ' tCntde odds ruio.-'Adjuated'for2 ge, number of live b'inhsi scliooling ,(+/=); and ynrs'since exposure to, eigarette smoke oeasediu the home.or'workp(aee.-3Fiom one or.bodi parents.-'From spouse„in-laws, children, orodhercoo-habinnts:-sFrom spouse, parents, in-laws, chili orocherco-habitantswhb smokedlat.h'ome in die.e presence of'the subject. By period in life TAiBt.E 111'.- MEASUREMENTS'~.OF PASSIVE~SMOK7NG AND~RR FOR LUNG.CA'NCER Totat years Exposure ~~s RRI (95% ~CD 1 R~Rt ~(95'4 ~~CI) ~ Eaposure 0 22140 1.00 1.00 0 l'-19 2'0/28' 1.30 (0~63; 3.6'8)' 1.95 (0.72; 5.31')' <I 20-34 24'139' 1.12'(0.59; 3.06), I'.36'(0:55, 3.36)' <2' 35+ 22y30, I_33'(0.79; 4'.44)' 2.26 (0:90. 5.67) 2+ Ad)usued'RR' ~(95 % ~ C)) , L 00 ~ Houmlday ~ RR' (95lL~ CU~ RR~'(957f~ Cl) 22140 1.00 1.00. 15/29 0.94 (0:4'1', 2:63) 1-U(0.37, 2'.94)' 330 11 1.94 (1.24', 6,74) 4,.10i(1.59, 10.61) 18137 0.88 (0;42: 2.42) 1100'(0.39, 2.58) TouJ hours (in handreda) Cigaremes/days" Eaposure controis RR'(9S'YSCI): RR='(951i,Cl). Eaposuce Ca~ COMro11 RR' (957.'CI) RR' (95IxCI). 0 22/40 1.00 1-W 0 25/48 1-00i 1.00 1-10 10 25/38 1.20 (0:60,3,.67) 1-68 (0.64',4,.45) 1-10 13116 1.56 (0.74, 4:96)' 1.83'(0.65, 5'.11') 101-200 23'/27' 1.55 (0:88; 5.53)' 2.28 (0.91. 5-72)' l 1-20 , 27/33 1.57 ; (1.00, 4.99) 2.56 (IP.06, 6.19) 20I1+ 18/32 1,.02'(0:.54, 3.47) 11.42 (0:56: 3.62) 21+ 23/40 1.10 (0:5i1, 2.47) 1i.21 (0.5'11, 2'.86) . Crude odCfs.ratio.-lAdjustad for ag4, number of live.births, schooling(+!-),.andlyears since exposurrto~cigarene smoke.cnsed in tlwbwneor workplace:-37lie sum ofnumberW.eigan:nes/diysmokedbyeaeh househ'oldimemberweighted by the years of exposure from tAarsource: MarRel-Haensut'unnd analysis:..Years: 0..55i'.)ours: 0.757 tioursiday: 0:70; cfg/day: 0.6T Liogisoc adjusted trend analysis: Years: 0.23,: hours: 0.98; hours/diyr.0.86i ,ciglday: .0 .~63'. TABLE M-EFFECTSOF 1NCREAS'ING:YEARSAND:MEAN HOURS)DAYOF TABLE V - EFFECTS~OFiNCREASING YEARS AND EA'RI]ER.AGE OF'INI1rIAL TOBACCO.EXPOS'URE EXPOSURE TO CIGARETTE SMOKE Yeanof ea posure Yearsof e:p osure 1-24 ?S+ 1-24 :S+ Mean hours per day©f exposure RRI' RR? R'RZ Age at first exposure RR? R.;RI RR2. < 1.5 11.33-1 2_224 1.47 2.13. >_ 25 l.501 L95`' 1L50 1.67 (19126)' (21l/26) (20Y25)3 (8y10). y ).5 1.02 1,21 1'.07 1.45' <24 1.00 1.35 1.25 1.86 (9116). ('17129) (8/95) (28Y42). 'Crude.odds ra and'years.since e tio:= Adjusted forage,.nu xposurc io eigarene.smoke c mber of'live eased in the births, home o schooling r+vorkplaoe: -' 9" CL 1.33 (0:68:. 4.00). 1-47' (0.74;.4.30):. Ii02(0.39. 3..45)., 1:07(0.5~7: 3.39): -'95% Cl: 2.22 (0.79;,6.21);,2:,13 (0:84- 5.43); 1,2I (0 37; 3.96). 1.45 (0:56. 3..78): -'Numb'er of' cases/eumber of eontrols: 22eases ,and. 40 controis's hsdd no exposure = RRI.00. 'tnde odds rario.= Adjustad forage, numbcr of'live.birrtAsl schooling (+/ ), ard yesn:since Kxposure tociguene smoke.oessed in the fiome or workyl.ce.-s95 x.. CI¢' Ii50 (0.2L 3.99):. 1:50'(0.47. 4;64), 1.00 (0;41. 3.42), 125'.(0.76. 3.60). -'95 xCl: 1-95'. (0.76; 4;M 1.67 ('DS2 ; 5-33). 1.35 ~ (0.30... 6.18), 1.86 (0.78, 4.46). 3Numtiec of . cuesl,n¢nber of wnaolf. 24 ases and 145 . controls had no

PASSME'SNfOKIti1G IN CHINESE FEMALES 1655 e..._ e.. .... ... r«r..,., ,.,. ..,.. M .... ~. .r.« _ .,. ... ~... ..~~ ,.~, r....,,.. u..a.. r.w. ~...,. FicuR>; l- bteasuretnentsforpassive smoking,and RR for lung can- cer. 'Adjusted i for age, number of live binhs; schooling (+J=) and years since exposure to cigarette:stnoke ceased in the home or work- place. 'p !~ 0:05,. the squamous or smaillcel[' lung tumors than' among the ade- nocarcinoma or large-cell types (Table VT). This was espe- cially true for'the adjusted RR in the former group,as 3 of the 4' measurements consistently indicated increasing nsk~ with increasing exposure. Location by lobe' Eighty of the cases had the main rumor residing in one of the Ibbes. The remaining 8 cases, with primary tumors in the right or'left main bronchus„or in the right'intermedius region, were too few for analysis. Calculations of,the RR showed that none of the crttdo or adjusted' values were significant for upper-lobe rumors (Table VII). For the middle or lower lobes, all of the adjusted RR were in the'comparatively, higher range of 1.9-3.5 for those with some passive exposure. Moreover, for 3' of the exposure measurements, total years, hours/day, and cigarettes/day, the'confdence intervals for the crude and! adjµsted RR indicated some borderline significant values. However, none of the trend' analyses for the lobe data came outsignilicantL T.VBLE M- MEICSUREMEwTS OF'PASSIVE SMOKING'AND RR FOR LUNG CANCER BY WSTOLOGICAL TYPE' ,.e,,.: o«..w. Squsmoioor small+eeR Adeoooreinarwor Iargecell Numberof cuesJ' Nym6er ofinusl number~of RR!~~(95!<~CI) RR; (95'x ':Cl)~. numberofRR'(95.% ~.C7)~. RR=(9S7f,C7)~. coards ~ aorwols 1.00 12/40 1.00 11.001 1.58 (0:37, 6:77) 17d46' 2.11 (0:54, 3'.74) 2.07 (0:64', 6:71) 1.82 (0:49, 6:80) 17/51 1.90 (0:51, 3.27) 11.43 (0:5i1L 4.02) 1.00 12140 1.00 11.00 1.40 (0:34', 5:77) 118J56 1.07 (0:48, 3:0'5) 1L70 (0.55. 5.20) 2:04'(0!53', 7:85) 16/41 1.30 (0.59, 4.02) 11.57 (0.55, 4.49) 1.00 12/40 1.00 IL00 1.34!(03 1, 5.84) 17/44 1.29 (0:56, 161) 2.19 (0:7'1, 6:77) 2.01 (0:521 7:72): 17/53 1.07 (0.49, 123) 1:34 (0:47, 3.82) 1.00 13'/48 1.00 1.00 2,02'(0.53', 7.68) l2/'26 1.70 (0:77, 5J2) 2.05 (0:63', 6.72) 1.19 (0.36; 3,.93)' 19/62 1.13 (0:59. 3.57) 1.88 (0':68; 5.17) 'CrWe odds ntio: -=Adjusced forage, number ofdive iiirtRs. se)iooling (+!-), and.yeansince exposuee tocigarene smoke eeuod in dx bome.orworY.placc. Total years 0 71'40 1.00 1-26' 10/46 l.2?t (0.37. 5:40) 27+ 15/5'li 1.68'(0.47, 5.79) Total hours (in hundreds) 01 7140 1.00 1-150 12/56 1.22',(0.34.,4.711) ISI+ 13/4I1 1.81 (0.52, 6'.54) Hours/day 0 7/40 1_W <'I.3! 8/44 1.04 (0.31. 4.70) _> 1,.3! 17/53' 1_83 (0.52. 6.69) Cigarettes/day. 0 9/48'. 1.00, 1-19 9/26 D.85 (0.57, 7.20) 20+ 14J6'2 1i?0 (0:36. 3.31) a TABLE' Vtl -WEA'SUREMENTSOF'PASSIVE SMOKING AND ~R~R FOR LUNGCANCER BY WBAR'UOCATION i Upper~lobes~ Middle or~.loWer lobes ~ Numi&nofcasey number otmnoolr RR'(95'7.~CI)~.. RRFf95~R'CI) VumbefiOlcases/ ' number of conrroli~. RR' (9515~.C1)~ RRz(95~4~CI) Total years 0 10740 1.00 1.00 L1 /40 1.00 1.00 t,-,6. 11/46 0.96 (0:43, 3.82) 0.98 (0!27. 3.64) 17146 1.34 (0,86; 8.72) 3.08 (0:83' 11.38) 27+ 161511 1.25 (0.40„2.87)~ 1.42 (0!46, 4.42) 15/51 1.07 (0:62: 6.15) 2.13 (0:62, 7.24) Total hours (in hundreds) 0 10/40 1.00 1.00 11/40 1.00 tL00 1'-1'50 15156 1.07 (0.30;,2!38) 1.30 (0:38, 4.50) 18156 1.17 (0.76; 7.26) 2:37 (0:67, 8.35) 151+ 12/411 1.17 (0_38„3'.0',l) 1_23'(0:39. 3.91) 14/41 1.24 (0_68; 7.17) 2.51 (01721 8.84) Hours/day 0 10/40 1_00 1.00 11/40 U.00 1.00 < I'.3 7144 0.64 (0.1'5, 1.58) 0.69 (0.18; 2.61) t7/44 1140(0,.95; 9.5 1) 3'.24 (0;90, 11.66). > 1'_3 20/53 1.51 (0:51, 330) 1.64 (0,.54, 5:01') 15/53 (L03 (0.55, 5.55) 1.97(0!57; 6.82) Cigarenes/day 0. 10/48 1!.00I IL00 12/48 1100 00 1 1-19. 10t26 1185 (0.57, 5:39) 2.32 (0:62, 8.76) 12R6 1'.85 (1L08, 10 39) . 3' 49 (0!981 12.50) 20+ 17J6'2 132 (0:48, 3.32) 1L79'(0.59; 5.45) 17/62 1'.10 (0,.61, . 4_61') . 1.93 (0j63, . 5.95) 'Cn'deodds ratio: =Adjustedforage, number, of'live~births, sctiooling.,(+/~-), and yesrssincetxposuretocigarette.smoke,ceLsed,in the.homeor workplace:..

166 KCDO ET A L. Pruzimallperipheral location Among the 85 determinable cases. 46 had'' peripheral tu- mors, and 39 proximal tumors. Although oniy the crude RR of 2.00 and adjusted RR of 3.52 for 1-19 cigarettes/day were slightly significant for the proximal tumors, in general. all of the crude and adjusted RR for the peripheral tumors were greater than 1.00 ('Ilable V'III). frftsrologicall type and location ln order to see whether any particular combination~of histo- logical type,, lbbe, or proximalYperipherall location of the tu- mor woul& result in stronger dose-response patttrns by the 4 118.00 14.00 Y 121.00 ~ 10.001 ~ lifetime measurements of passive smoking, RR were analyzed : for the 12 possible 1:1 combinations. We were unable to segregate the : cases into any finer categories than 2 of the 3 groups because of the small resulting number of cases for analysis. Space does not allow us to present all the tables; but the best combination was that of peripheral tumors in the 0 0: 6.00 Total! years Total hours HoursJday middle or lower lobes (Table IX). Among the RR, signiftcant' a. 00 ~ or nearly significant figures were found for the crude or adjusted RR" relating,to at least one of the exposure categories for each type of measurement. Moreover, the adjusted RR tended to range between: the relatively high values of 6,5 to .18.7 for those withisortte exposure (Fig. 2)„ and most of~these were significant or' nearly significant. None of the trend'tests came out significant, but this and the tendency, for the higher levels of exposure to have lower RR than the low levels of 2.00 exposure may' have been due to the small number of' cases' Ftcuue 2- Measurements of passivrsmoking;and RR for peripheral (N -24),, lung cancers in the middle or lower lobes. Adjusted odds ratio. Although not as apparent, squamous and small~cell Iung cancers in the middle or lower Ibbes (Fig. 3) also seemed to show some positive association with passi ve smoking: There were only 18I cases with ~ this type for analysis and, none of the 8.00 I+tR' or tests for trend were found 1 to be statistically significano 7.00 (Table X). Yet it was promising to see that all the RR with some exposure were greater than 1.0. Among the highest' exposure levels for the adjusted RR, values as high as 7,0 were ftiund for totallhours, and 6:2 for hours/day. DI'SOUSSIOM' For comparative purposes, the more commonly used, mea- surements of' passive smoking based on yes/no questions of whether household co-habitants (husband; childhood/adtllt- hood; or others) had'smoked, or on the number, of cigarettes the husband smokedI per,day„ were presentedL Only'the crude. RR of 2.37 (95% CI:1.03-5.94) for husbands smoking 1',10 cigarettes/day was of borderline significance and none:of the: adjusted odds ratios were significant at the <5% probability level. There was little indicationithat increasing,levels of such exposure led to increased ItRL On the basis of our extensive life-history data, we were able to calculate the total years, hours, mean hours/day, and ciga- rettesAday to which the subjpcts had ibeen exposed to tobacco smoke at home or at, work. Our estimates were based on the understanding that the household composition of each subject would'ehange as she progressed through the life-cycle of~birth, childhood, adulthood, marriage, motherhood and, for 27'!a; widowhood. We also included exposures fromieach workplace at which the subject had worked itir at least 3' montlis. In our adjusted RR, the effect of cessation of exposure to passive smoking was accounted forby putting in the years thaCexpo• sure had ceasedl an home and/or workplace as a continuous regressor variable. Despite such detailed accounting, we were unable to find a significant trend in the crude or adjustedi RR for these 4 lifetime measurements of' passive smoking. Although the RR for the intermediate level exposures of, hours/day and ciga- 6. 00 2:. 00 0 f! L , None Low Exposure Levels ~- til o n e Low Total hours Highi Exposure Levels Fiouru=3 - Measurements of passivesmoking and RR forsquarnous and small-cell lung,cancer in the middle or lower lobes. Adjusted odds ratio. rettes/day were sigttificanr, the RR at the highest levels of exposure for these two variables fell to a non-significant 1.0- .t 1.2. In fact, the RR for the highest , exposure levels for 3 out of the 4' measurements were below all of those with lower ~ ' exposures, andd ranged from a very weak lL0 to: 1.4 . On the other hand, most of the crude, and adjusted RR were greater : than 1.0@..

PASSIVE Sb(OKING'IN CHINESE FEMALES' TABLE'~VIII - MEASUREMEMTS~.OF PA'SSIwE'~S6IOKING:AND~RRFOR'LUNG CANCERBY'L'OCATIONOF'7R1MOR~~ 167' Perpheral. Procimal NumEcr o( cnus/. number of contrds . RR"(9'SR CI) RR, ''(957F CI) NunOef'o( cua/ numDer.of controls RR' (95R CI) RRJ` (95'R'CI) Total' years 0. 10/40' 11.00, IL00 11/40 1.00 1.00 1-26 18Y46 1 L57 (0.59, 4:84) 1 L52 (0.44. 5.17), 14/46 1.11 (0.50; ,4'.14) i 2:15 (0.64; 7:19) 27+ 18151 I L41 (0.64, 4.78) 1 l84 (0.62, 5.45) 1 14/5II i 1.00 (0.43,:3:51)' 1.58'(0.51, 4.92) Toral hours . (in hundreds) 01 10140 1.00 1.00 1 v40 1.00 1.00 1-1'50' 20156 1.43 (0.63; 4.97) 1.82(U7; 5.85) 16/56 1.04 (0.46, 3.53) 1.86'(0.58 i 5.97). 151+ 1614111 1.56 (0:60, 4.71) 1.66 (0:54: 5:06) 12141 t.06 1(0.47, 4,19)' 1.72'(0.54. 5_51) Hours/day 0 10/40 1.00 1.00 11140 1_00' t.oo. < 1.3 14/tl41 1.27 (0:56 . 4.62) 1.66 (0:52; 5.33) 13/44 1.07 (0.48, 3.94) 2.21 (0.63, 7.73). ;?t 13 22153' 1.66 (0.66; 4.98). 1.77 (0.59 , 5.32) 15/53 0.89 (0.44'. 3.69) 115'9 1(0.51, 4.93) Cigarettes/day 0 12r48 1.00 1.00 12/48'. 1'.00' li.00' 1-d9 1 Id26' 1.69 (0.73: 6'.14) 1.91 (0.57, 6.35) l3V26 2.00 (0.98; 9.17) 3.52 (1.01, 12.27) 20+ 23/62', 1.48I(0.70; 4.34) ' 1_79 (0.64: 5.03) 12;162 0.77 (0.34. 2.45)-'- ri23 (0.42: 3.62)" 'Cn+de odds rario.?Adjusted for age, number of live (+irths, sclrooling (+/-), and years siooe eaposure to,aguette smoke aeued in ttw hotne,or .wrkplaon. TABLE IX - MEA:SUREMENT6 OF PASSIVE SMOKING AND RR FOR PERIPHERAL LUNG CANCERS IN'THE MIDDLE OR'LOM'FJt LOBES' Exposure .yumtier~af aaset/ number of eontrols RR' ro5sCn RR=(95x Cn Total years 0 4140 1.00' 1.00 1-26 10/46 2.17 (0.98, 84'.95)' 10:44I(0.91, 1119-53), 27+ 10/5',1i 1.96 (0.88, 66:91)i 8!61 (0.84: 88.21') Total hours (in hundreds) 0 4/40 1.00 1.00 1-150 12/56 2:J4 (1.24,,110.,17) 13:51 (1Lll6,,157.74) M + 8/411 1.95'(0.69, 56.35) 7.02 (0.64, 76_93) Hours/day 0. 4/40 1.00 . 1.00 < 1.3' _> 1.3 ' 11144 9/53 2.50 (1.711', 160.18) 1.70 (0:62, 49:89) 1I8.70 (1'.53. 228!03)6.49'(0.60. 70:37) Cigarettes/day. 01 6/48 1.00 1.00 1-19 6/26 1.85'(0!95', 24.36) 5.53 (0.79, 38:86) 20+ I2/62 L.55'(0:74, 13.14) 4-16 (0:77. 22-55) 'Crude odds ratio:-lAdjusted for age, number ofil74ebirtht. sdtooling.(+h-). and years since..ezpnsure to:cigarene smoke ceased in tlie.liomee orr workptaee: , Mintel-Haenszel trend analysis: Years: 0.15t houra: 0. 16: hours/day: 0. 14: cigfday: 0.29: Logistic adjusted trend analysis: Years: 0.15; hours:0.66; hours/day: 0.53':'uglday:'0R2. TABLE ~~X:- MEASUREM~ENTS~ OF PASSIVE SMOKING A'ND~.RR~, FOR SQUAMOUSAND SMALL-CELL LUNG ~CANCERSIIN THE'~MIDDLE OR LOwER~ LOBES ~~. Ezposurc Numdcrof iaxsJ number of controls RR"(95 % ~ CD i RR= (95"b'. Cq Total year6'. 0 3/40 1.00 11.00 1-26 7146 2.03 (0.52, 44'.44)' 5.29 (0:5i1, 54:71) 27+ 8/51' 2:09 (0.42,,33101) 3.97 (0:4i1, 38.22) Total hours (in hundreds) 0 3/40 1.00 1.00 1-150 6/56' 1.43 (0:35„ 29.32) 3'.44 (0:35, 34.17) 15 i1 + 9/41 2.93 (0.59, 46'.98) 7'.01' (0:64, 76_60) Hours/day 0 3/40 1.00 1.00 <,1.3' 4/44 1.21 (0.30, 29!64) 3105 (0.281 33.14) ~t 1.3' 1'1V53 2.77 (0:57, 44:05) 6:16 (0.59;,64.48) Cigarettes/day 0' 4148 t.00 1.00 1-19 5/26 2.31 (0:58', 23.25) 3:97 (0,.54„29.30) 20+ 9162 U.74 (0.44, 11.87) 2.58'(0.42 15.93) 'Crude odds ratio.?Adjusted for agC, number of lirelfirths„sohooling.(+d-);,and years since exposure to oigarette smokcc ceased iin tlie:liome orworkp,taoe;. Mintel-Hxnszei trend analyRis:!Yars- 0.23; hours: 0.20; Aours/day:'0:26; cig/day,: 0.20! Logistic adjusted trend analypis: Years: 0.71; hours: 0:76; hours/day: 0.70: cig/day,: 0.7& .

168 Measurements based on increasing intensity of exposure; combined analyses showing some promise, perhaps the best ' defined as increasing years (or hours: or cigartttes/day) by RR' would have been obtained if analysis: had been done with mean hours/day of exposure, also di& not indicate a dose- squamous or small-cell peripheral' tumors in the middle or response relationship. Likewise, the analysis of'total years of lower lobes. We were unable to do these calculations because ' exposure withi age of exposure did not suggest that earlier age only 8'cases fitted into this category. of initial exposure and increasing years of exposure led to Actually, the finding of a possiblb risk of squamous and'& higher RR. It was troubling to find that in both types of small-cell tumors in the middle or lower lobes was somewhat analysis, the RR' for the lowest amounts of exposure were unexpected, given that dust particles tend to adhere to the among the highest values. upper lobes, and tuberculosis usually affects the upper lobes. Dalhamn eral' (1968) noted from their study of the retention To see whether calcified foci or fibrosis in the upper lobes of cigarette smoke components in human lung, that water- could account for the higher RR in the middle or lower lobes insoluble volatile compounds and particulate matter from.cig- because the previous presence of such lesions might disturb arettes tended to be deposited primarily in the dteper parts of d+e expected distribution of inhaled particulate or gaseous the respiratory tree. Since adenocarcinoma is predominant matter;, most of the chest radiographs of cases,with squamous among non-smoker lung cancer cases (59 % of our typed cases) and small-cell lung tumors were re-examined. No significant and, it is generally a peripheral tumor, we wanted to see difference was foundl in the proportion of positive cases with whether the passive smoking measurements would exhibit a upper lobe vs, loM1Ver lobe tumors. more consistent' pattern among the adenocarcinoma and, large- In our analysis of all'never-smoked cases, the lack of a dose- cell types, andlor among the peripheral tumors, In general, response pattern, and an i almost consistent drop in the RR at' the peripheral' tumors as a group showed stronger dbse-re• the higheste doses of exposure would seem to lend little, or spottse results thanthe,adenocarcinomas._ ,. .._y only weak-suppont for thcpassiwe-smoking linkage with lung The RR for total years, hours, and hours/day measurements, cancer for women in Hong Kong. This>might be due to the of'squamous and smallicelHung rumors indicated consistently fact that it has been estimated (Rylandbr eral:, 1983)ithat the elevated I risks with increasing exposure. This pattern was not non-smoker exposed to environmental tobacco smoke receives foun& for any of the adjusted RR for adenocarcinorna or large- about 1% of the active smoker's dose of tobacco smoke based cell lung cancers. This association of histology with passive on cotin'tnc levels in the body, and this is' roughly equivalent smoking is also sugmested'from previous studies by Trichopou- to the tobacco smoke of 0.1-I.0 cigarette inhaled by an active los et al. (1981I) and Correa et al. (1983)~ smoker in a day. Moreover, a 15- to 17-year longitudinal study Analysis of the cases by the lobe location of the tumor was of 97' non-smoking females in Holland did not find an associ- done to see whether the primary tumor resided more fre- ation between passive smoking exposure and pulmonary func- quently in the upper lobes than in the lower lobes. This is tion decline (Brunekreef a al:, 1985). Thus the effects of because it is known that when dust is inhaled, it' f rsn enters passive smoking might be so weak that they are easily over- the upper lobes where much of it is deposited, and then travels shadowed by other environmental factors such as diecor ex- down to the lower lobes (Time, 1980). Furthermore, it has posure to inhaled gaseouslparticuiatematterfrom other sources been observed (J.Nl-C. Ho, personal observation) that up to in the home or the workplace. half of the Hong Kong adult population have radiologically When the lung tumors were scgregated by histological type evident scars on the upper lobes of theiriungs. Most of these and location~ the resulting analyses show.ed', that peripheral scars are due to previous tuberculosis infection. Since "lung, rumors in the middle or lower lobes, and squamous or small- cancer is more common in the scarred and chronically diseased' cell tumors in the same lobes; exhibited better RR patterns for lung" (Stone et aL, 1978), we were interested to see whether passive smoking: ini terms of consistency, strength, andldose- the lobe data would substantiate any of these possibilities. In response. We are not sure whether this proclivity for passive~ fact, 37, of the lung,cancers were found in the upper lobes, and smoking-related lung, tumors to reside in the middle or lower 43 in the middle or lower lobes. The results from the RJZ' lobes might be due to the fact that: the lower lobes have more estimates from the 4'types of measurements didinot show the bronchial cells at risk than,the upper lobes, or whether thee upper lobes to be more sensitive to environmental tobacco size; weight, or composition of gaseous or particulate matter smoke, from passive smoking, may favor its adherence to the periph- Wynder an&Goodman (1983) suggested that lung cancer in eral areas and the lower lobes. Nevertheless, the overall pro- ,on-smokers.was more likely to occur in the periphery of'the portion of lung tumors in the middle or lower lobes among our ~ ` .ung. This was found in our study, as 54 `90 of the determinable 88 cases ranged' from 27% for the peripheral tumors to 20% cases had peripheral' tumors vs. 4696 with proximal tumors. for the squamous or smalf«cell' tumors. Thus, the majority of I Moreover, the pattern of RR' with the various measurements lung cancers among our non-smoking pwpulation,were proba- ~ of passivesmolung indicated that penipheral'tumors seemed to bly due to some factor(s) which yet remain to be idendfiedl exhibit better dose-respottsc RR than,proximal tumors. The results from this study;,show,ing a weak effect of passive When the RR' were calculated for the 12 possible U:l com- smoking on the risk of lung cancer among never-smoked Hong binations resulting from histological type, location by lobe, or Kong Chinese women, mustbe interpreted cautiousdy,,sinceit proximal7peripheral' tumors, the highest RIt were foundi for was based',on only 88 cases and 137 controls. With this sample peripheral tumors in the middle or lower lobes. Significant size, RR less than approximately 1.4 would be difficult to c on near y s~gpt other studies is therefore needed.. M- . -.- . e t measurements: Although RR at the lower doses tended to be p gr~ea r w en e cases were strau oe higher, than that for the higher doses, the data were consistent by' histological type and location of the primary tumor. How- in that all' the RR for those with some exposure were much ever, these data seem consistent with the findings from other greater than 1.0, and the adjusted' RR for at leasb one of the epidemiological, biochemical, and physiological studies in RR for each, type of ineasuremenn was statistically significant showing, higher risks for squamous-cell turttors in the periph- eral areas of the lung:. Confumation of these findings from ant ! . .fi adjusted Rft' as' high as 18:7' were found for some of these deteet wtdt 95% power an& ar the 5%, level of s3gnificance: ; This roblerrt was even t h h 'l' d showing some dose-response pattern.. With the above two versity of Hong,Kong for financial',assistance ia the cazrymg s~: than the others with total hours and hours/day measurements We thank the Hong Kong'Anti-Cancer Society and the Utu* :~ The RR analysis for squatnous and small-cell lung cancers in uhemiddle or lower lobes ~ alsoappaaredi sornewhat' better AMNoWLEmGtMEtr,~~ . . , .. . ... .. _ .. . ,. . . _.... ~'. 'k '.; . 1 k - " . .- ~ ' ~ ~ f

W3~ ~~ .._.~..,... ~. ._. ~.. .~. i ~, PASSIVE SMOKING IN CHINESE FENIALES' 169 out of tho fieldwork. We are also indebted to the US National the data. The secretareal l assistance of Mrs. T. Lain, Ms. A. Cancer Institute's Fogarty International Center for sponsoring Chow and Ms. M. Chi„and the graphics work of the Medical a 4imonth Visiting Scientist post in the Epidemiology and Illustration Unit, are gratefully'acknowledged. Biostatistics Program. which was invaluable in the anaHysis of REFERENCES BRFSt.ow; N.E.. and DAY, N:E., Statistical merluada in cancervesearch, IARC Scicntific,Publication 32. IARC, Lyon (1980). BRUNEKAEEF, B.,,FI.Wl{ER„P...REMLN. Bl, VAN DER'LENDE. R., ScHou. hrt,1;. and QqhnuEtt: P.. Ittdbor air pollution and',its effecoon pulmonary function of adhlt non-stnoking wottten. IIIl Passive smoking and!pultno- nary function. !nL J., Epident.. 14. 227=230 (1985). CH,w, W.C., and,FuNa: S.C., Lung caneer'in non-smokers in Hong Kong. In: E. Grundtnann (ed:),, Cancer campoign, Vol. 6, Cancer epide- miology pp. 199-202, G:,Fischer, Stuttgart. New York (1982). CowtEA, P.. PtcKLE. L.W.. FoNrH..M: E., LrN. Y.. and HxENSiEL. W., Passive smoking and lung cancer: Lancet. II, 595-597 (1983). DAutAntN, T.. EnFotcs: M-L., and RYUnoEn. R., Retention of cigarette smoke components in human lungs. Arch: environ. fdlth:, 17, 746-748 (11968). FateDUUN. G:D:, P6rrrn, ID,.B:, and BAyQt., R.ID:, Prevalence and cor- relates of passive smoking;,.tmer. J.pttg ,RIth', 73, 401-405 (11983), Koo. L.C.. Ho. J.H-C., and LEE.. N.,,An analysis of some risk factors for lung cancer in Hong Kong: Inr. J.' Cancer. 35, 149-155'(J985). Koo. L.C:, Ho. J:H-C.. and Sn,w; D:, Active and Ipassive smoking among female lung earteer patients and controls in,Hong Kong. J. esp. din. Cancer Res., 4. 367-375 (1983). Koo. L.C:, Ho. J.H-C., and Sww. D.. Is passive smoking an added risk factor for lung cancer in Chinese women?'J. exp: cfin; Cancer Res.., 3, 277-283 (1984). LuaiN, JiH., A,computer program for the analysis of matched case-control studies. Comp. Biomed: Res., ,1A, ,138-143 (1981): MtLLER, G.H., Cancer, passive smoking and tnnemployed and employed . wives. Wesr.J.' Med:, 140, 632-635 (I1984)1 RYLANDER:.R.,.PEaEasoN, Y:, andSNEULx.Nt-C_, (eds:). EZSl-Environ+ mental robacco srrtoke. Atar; Geneva (1983). SANOeER; D:P., Wn.cox, A'.J., and EvEasoNl R:B., CumulaRve effects ' of'lifetime passive smoking on cancer'riskr lcncer; 1, 312-315 (1985). GARFINKEL, L.. Time trends in lung cancer mortaliry, among nonsmokers STONE. D.1:, GLUCKs M.C.. and PArttKH, N.K., Practicof poinrs in ptd= and a inote on patsive smolting. J nor. C'cncerlnsL„66, 1061-1066 (1981). monary diseases, p. 125. Medical Examination Publishing Company, New GARFINKEL. L., AuEttBACrt.,0., and JOUeERT.,(:.. Involuntary smoking Yordt(1978). and lung t:ancerr a case-cotttrol study,. J.' not. Cancer lnsr. , 75, 463-469 TIME. Atlas ojtfic body, p. 59. Rand McNally. Chicago (1980).. (1985). TpjaHOnouuos, D:. KAtANnror. A.. SPARROSi L., and MwMMHON.,B'., HtwArAMr; T., Non-smoking,wivcs of heavy smokers have a higher risk Lung;cancer and passive smoking. Inr. J. Cancer;,27, 14 (1981)- oflungcancer: a studydromJapan: Brir. med. J:.Z82, 183-185'(1981). Wu, A.H.,,HENOERSON, BIE.. PIKE. M.C.. and'Yua M.C., Smoking,and KAewT, G.C., and WYNDER. E.L.,, Lung, cancer in non+smokers. other risk factors for lung cancer in women. J: nat. Cmtcerltisr., 74, 747- Cancer. 53, 1214-1221 (11984)j 751 (1985). KNOTH, A., BONN, H., and SCHMiDr. F., Passiwrauchen als Lungenkrebs- WYNDER„ E!L., and GOODMAN, M.T.. Smoking and lung cancer: some ursache bei Nichtraucherinnen. Med. kJin. Pra.r:, 78, 54,59 (1983). unresolved issues. Epidem. Rev., 5,,177-207 (1983).