Tobacco Documents Online

i )urnai ul L pic^emioiogy VoC 23, No. 2 ©lnternationad Epidemiolog cal Association 1994 Printed in Great Britain Childhood and Adolescent Passive Smoking and the Risk of Female Lung Cancer FU-LIN WANG,' EDGAR JOHN LOVE,* N9NG LIUt'AND XU-DONG DAl1 Wang F-L (Department of Community Health Sciences„Faculty of Medicine; University of Calgary, 3330 HospitallDr. NW, Calgery, Alberta, Canada T2N 4N1)„Love EJ, Liu N and Dai X-d; Childhood and adolescent passive smoking, and the risk of female lung cancer. International Journal'oL Epidemiology 1994; 23: 223-230. Background. Few studies have reported4he relationship between passive smoking (PS) in early life and the risk of lung cancer. This study was done to evaluate the risk of female lung,cancer from PS, especially that during childhood', and adolescence. Methods. Using household exposure to tobacco smoke as an estimate of PS;,a 1:1i paired'case-control'study was conducted in Harbin; China, We interviewed 114 female primary lung cancer cases, aged 30-69 years„and their hospital+based'contfols. The controls were non.cancer patients, selected from the same hospital as the cases, an6 matched on age (± 5 years), residentiallarea and smoking status over, their Ilfetime: There were 59 pairs who ever smoked and 55 pairs who never smoked. Information on PS was collected by residence for each of the following, peri-+s: 0--6, 7-14, 15-r22, 23-30 and 31-69 years. Re: s: Household PS significantly increases the risk of female lung caneerfonthose exposed aVages 22 or younger, who have ever smoked. The risk was also increased for those non-smoking pairs when exposed under the age of 15 years. Exposure to maternal smoking at ages 14 or younger increased the risk by about 1170% (odds ratio, OR 2.7, 95% confidence interval [CI]: 1.49-4'.88); but not to paternal smoking (OR 1.4'0„95%;Cl: 0.92-2:50). The risk was highesTforthose exposed und'er the age of seven (OR 3.46, 95%CI: 1.80-6.65) and was also significant at ages 7-14 (OR 3!08', 95%CI: 1.62-5.57)landi 15-22' (OR', 3.10, 95%Cl: 1.52-6.31) years. Under the age of 23 years,, the OR increased with iamount ofiPS(P< 0.001). Of note, the OR'in all five exposure periods fornon-smoking pairs were similhr to those for all 1114 pairs studied. Conclusions. Household' PS, particularly that during childhood, increases the risk of female lung cancer. The assessment of PS'should be done by difierent'periodsof exposure. There is increasing evidence suggesting that involuntary or passive smoking (PS) increases the risk of lungcan- cer.l-6 However, the findings have not been consistent witKrespect to who are at higher risk: adults or children. Some studies have noted'the importance of exposure to tobacco smoke in early life in relation to cancer risk in aduli',ood,i-10' while others found no association for non rioking women exposed to tobaceo smoke during childhoo&a' In addition, themet'hodsto estimate PS'variedbyS study. The estimate of PS from only the husband or the spouse, which has often been used, is in question.1l-1. This study, using household exposure to tobacco smoke as an estimate of PS, attempted to clarify further the relationshiipbetweenfemalelung cancer an6 PS, ~DepartmenuofCommunity Health Sciences; FacultyofMedicine, Urti- Versity of Calgary; 3330 Hospital Dr. W, Calgary, Alberta, Canada T2N4N1. tDcpartment,of Epidemiology, Heilongjiang,lnstitute for Cancer Re- seareh~ Harbin„The People'sR~epublie ofChina especially that during childhood. Data were collectedlin the city of Harbin, the People's Republic of China, during 1985-11987. 223 MATERIALS AND METHODS The cases were female primary lung cancer patients who were selected from the Harbin Cancer~ Registry. The cniteria for inclusion were::Resident of H'arbin city, aged 30-69 years and diagnosed by histopathology during January 1985 and November 1986. A total of 114 cases were eligible and alllwer~e interviewed face-to-face: They includedl 55 adenocarcinoma„ 28' squamous cell carci- noma, 20 small cell and oat cell carcinoma, and l 1' other types. A reporting system was set up to, ensure quick registration of cases, and immediate investigation. The controls werepatients, withoutcancer„from ad- missions to the same hospital as the cases and matched on age (± 5 years), residential area and smoking status h over lifetime (whether or not a~ currenC or ex-smoker). a Inforrnation on sex„age, residentiallarea and thediagrao- N sis was obtained from the patient f ile and the preliminary (~ ~ ~ W ~ N

224 1N"TERNATIO*1AL 1OURNAL OF EPIDENUOLOGY T,ke1E I Ch'araeteruuca of rh'e studr subjects in a 1' 1, nratch'['d case-control study, Harbin. China. 1984-1957. Characteristics Cases No % Controls hJo~ % P-values' ADemographicCharacterustics Nationality Han 107' 93.9 112 98.2 0.174 Other 7' S.1 2 1.8 Ethnicity Heilongjiang, 57 50.0 49 43.0 Liaoning & 1i11n 20 17.5' 18 15.8 0;584 Shandon 22 19.3 29 25.4 Other 15 13.2 18 15.8 Education Literacy 59 51.8 65 57.0 Primary school 20 17.5 22 19.3 0,684 Middle se hool 27, 23.7' 20, I7,5 > College 8 7.0: 7 6.1 OccupationV1Chite collar 9 7.9' 5~ 4.4 Blue collar 63 5543'. 71 64.0 ~ 0-4'56 ~ Housewife 37 32:A'. 33 29.0 Other 5 4.4 1 2.6 Marital status Married IW 87:T 103', 90 4 0.546 VI'idowed/divorced 14 123 '. 11 9:6 B. Smoking Habits Ever smoked cigarettes Yes 59 51.8 59 51.8' 0.999 No 55 48.2 55 48;2' Age began smoking(h = 59) Mean ± SDb 19.1 ± 9.56 2].6 ± 10.0 0.179 Minimum tomaximum 5-59 6-58' Number of cigarettes/day (m=S9). Mean'SD 10.4 + 7.72 10.2 16.42' 0.013 Minimum to mavmum 3-34: 2-25 Years of smoking (n = 59) Mean + SD, 32.6 ± 13.2 31.1 ±' 12.6 0 545 Minimum to maximum 3-50 7-46 ' X' test for the differences in the percentage ofAemographic variables between cases and controls or Student's t test for the differences in means of, smoking habits between cases and eontrols. b Standard deviation. 'eligible' patients (i,e. age and residential area matched)l were ascertained. The first eligible patientwho was found to match~ the cancer patient for smoking status over lifetime was enrolled inithe study. The distribution of the controls was as follows:. li8%a had cardiovascular disease, 1''5%a had digestive system disease,. 22%had respiratory infections, 30% were in hospital for acute surgery', 10% had gynaecological problems, and 5% had other non-malignant disease. Using a pre-set questionnaire, the cases and controls were interviewed'in.hospitaTs or at the patient's home by trained interviewers; some of the cases and all the controls were interviewed by WF- The questionnaire included' questions on demographic information, resi- dential history; smoking history, and the ind'oor smoking habits of the regular fami.ly members, etc. Information on indoor smoking was~collected for each residence in which the subject had lived for 3 or more years. In this way, information on PS during different periods was also obtained: r

PASSIVE StvtOKING,AND THE RI6t<OF FEMALE LUNG CANCER 225, TAeLE 2 The distribution of sources of sidestream smoke by exposure period 'ur Harbin. China Study Nb. of'subject d s Sources ofsidestream smoke (%)' subjects expose Father Mother Grandparents Others Total A. Exposure at age 0 Cases 14 years 1041 45.2 38;5. 6 7 9.6 100.101 Controls 74~ 51.2 26.4 9.3 13 2 100!0 I B'. Exposure at age 15-69 years Husband Others Total Cases 85~ 69.4 30.6 100.0 Controls W 68.8 311.2 100.0 ' Number exposed to specifi¢ source/total exposures expressed as percentage. In this,study, PS was d'efined as 1 gramlday or more of indoor smoking, byanymember of the family sharing the same living accommodation as the subject for I year or more; a grarru!day is roughly equivalent to one ciga, rettlP or one pipe of tobacco per day. The cumulativee amt,..nti of PS was estimated using the weighted mean. Similarly, we also calculated the weighted mean of PS when the smoking status of the family member(s) changed. To measure the cumulative effect of PS on the cancer risk, a passive smoking index ('PSI) in grams of tobacco x yearsof smoking, was calculated, i.e. PSI = the average daily consumption of tobacco (grams)~byfamily members.x years of smoking Case-control pairs were classified into the five age groups of exposure: 0-6, 7-14, 15-22, 23'-30 and 30- 69 years. Of the 114 case-control pairs, there are 59 pairs who ever smoked and 55 pairs who never smoked. McNemar's analysis13 was used~ throughout for calculating odds ratio (OR). The 95%confid'ence interval (CI),for OR wascalculated using Cornfield's method. The xZ statistic was~used to test for trend.The analw-;is was done on a personal computer using Epi infl ;ersion 5). RESULTS The cases and controls were comparable with respect to the matching variables (age, residential area within Harbin and smokingstatusover lifethme) and the follow- ing: nationality, ethnic origin, education and occupational'statvs (last job), the number of years that they have lived in the cityof H'arbin and previous history of respiratory disease, i.e. chronic bronchitis, emphy sema or pulmonary tuberculosis: Table I summarizes the selected demographic characteristics and smoking habits of the cases and controls. For the 59 smoking pairs, there are no differences between the cases and controls in age at which smoking began and years of smoking,, However, cases smoked more cigarettes per day than controls. Table 2 presents the distribution of sources of side- stream smoke for two age groups: 0-14 years and -- 15 years. In the f rst~ group, although the sidestream smoke exposure was mainly fromiparents, the percentage from mothers for the cases (38'.5%)! was higher than that for controls (2b.4% o). For the older group, husbands became the most important source of sidestream smoke, ac- counting for more than two-thirds for both casesan& controls. TxeLE 3 Risk offemale lung cancer during childhood (Q-14 } ears 1- exposure toraternal andnwternal,tohacco smoking in Harbin, China E Number and proportion exposed! xposure Cases Controls Nb. °/ No, Paternal 47' 41.2 38 smoking Maternall 40' 35.I 19 smoking Odds ratio 95% P-value confidence ia interval 33.3 1.40 0.79-2.50 0273 16.7 170 1.39--5 30 0.003

226 lNTER~~N~A~T1OtiA~L'. JOL'R`:~A~L OF EPIDESIIK?LOGY~ TaBLE~4 Risk offeniale (ung~,cance°rfirrm hot.~chuldci-poazre^e to~tobacco smok~e by~~e1P«sure~prriod~ (age) in Fdarbin. Chtna~ Exposure i d Number and proportion exposed per o (age in vears) Cases No: "~ Controls s No. Oddsrauo' 95% P-valiie confidence interval 0-6 75 65.8 48 42. I 3.46 1.80-6.65 < 0.001, 7-14 73 64.0 48 42,1 3.08 1.62-5.87 < 0.001 I5-22 85 74.6 64 56.1 3:10 1 1.52-6.31 < 0.002 23-30 70 67'-4 72 63 2 0:91 0?8-2.98 0-647 39-69' 75 65.8 80 70.2 0278 0':36-11.69 0.348' ' OR'from McNema~r, analysis;,OR = b/c, x' = (b - c - 1)'~/jb + c)', As shown in Table 3, the difference in percentage of exposure to maternal smoking under the age of 15 be- tween cases and controls was highly significant (P = 0!003); the OR for female lung,cancer associated with exposure to maternal smoking was 2:70 (95% CL• 1.39- 5:30). In contrast, the difference between childhood exposure to paternal smokingwas notstatisticallysignif, icaa (,P= 0.273).Table 4 shows the OR for lung cancer when PS oc- curred for the following age groups: 0-6, 7-14, 15-22, 23-30 years and 31-69'years. It was indicated'that the risk for lhng cancer was highest in those exposedlunder the age of l years (OR 3.46, 95% CI: 11.80-6.65) and wass also significant at~ ages 7-14 (OR 3.08', 95% Cl: 1.62- 5.57) and 15-22'(OR 3:10, 95% CI: 1.52-6.31)': Analysis of lifetime exposure (PS occurring at any time) showed that thecrudeOR for lung cancer associated with,lifetime exposure was2.67 (95%oCI: 0.90-8.88), which was close to the significant level (P= 0.055). In: reviewing these results, itishould be remembered that most children in the People's Republic of China remain at home until'age 7 when they start to atrttend school on a regular basis. TABLE 5 Female lung,cancer risk from household PS, by exposure period'i» smoking pairs and, non-smokingpairs. Harbin. China Exposure Number and proportion exposed eriod p (agein Cases years) No. % Controls No. % Odds ratio' 95% confidence interYal P=vallre A. Smoking pairs (n = 59) 0-6 47 79.7 33 55.9 3:33 1.17-10.76 0.011 7-14: 47, 79:7' 34: 57.6 2:86 1.06-8.43 0.021 15-22 53 89:8' 42 71.2 4.67 1.12-29.42 0.015 23-30 4,1 69.5 42 71.2 0!89 0:28-2.82 0:999 31-69 43 72.9 47, 79.7 0.67 0.22-1.92 0:502 B. Non-smoking pairs (n = 55) ~ M 28' 50:9' 15 27 3 3 60 1 15-13' 33' 0:012 N 7-14 26 47.3 14 . 25.5 . 3.40 . . 1.08'-12'69 0!009' M W 15-22 32 58,2' 22' 40.0 2.43 0:88'-7:33 0!066 ~ 23-30 29 52:7 30~ 54.6 0.93 0.38-2.25 0.999 W 31-69 32 58:2' 31 60.0 0.91 0:32-2:53' 0.999 ' OR from McNemar analysis. 'Wn V'

PASSIVE SMOKING AND THE RISI:OF FEMALE LUNG CANCER TABLE 6 Risk of feniale hung cancer frorn PS b1krel of e.rposure and period of e.iposure in Harhin. China Period of Level of Cases Controlk 95°b exposure exposure (age in years) ~ (grartilday) No, % of No. t'otal %of total Odds ratio confidence interval P-value < 5 54 47.3 82' 719 1.00 0-6 5-14 35 30.7 21 18.4 2.53' 1.27-5.06 0:007 . 15 25 22.0 11 9.7 3:45 i .48-8,20 0!003 Test for trend P<0.001 < 5 51 44.7 81 71.1 1.00! 7-14 5-14 37' 32.5 21 18.4 2.80, 1!.41-5.59 0.002 > 15 26 22.8 12 10.5 3.44 1.50-7.99 0.002 Test for trend P<0!001 <5 42~ 36.8 65 57.0 V00 15-22 5r141 44 3816 38 33:3 11.79 0.96-3.35 0.068 = 15, 28 24.6 Il. 9.7 3.94 1.66-9.49 < 0.001' Test~~for trend'~ P < 0.0011 < 5 64 56:1 64 56.1 1.00 23-30 5-14 39 34.2 40 35:1 0.98 0;54-1.78 0:957, . 15 1111 9.7 10 8:8 1.10 0!40-3!04 0:974' Test for trend P = 0.871 < 5 56 49.1 58' 50 9 1.00 31-69 5-14 47 41.2 45 39_5' 1.08 0.60=I.95 0j889' ~15 ll 9.7 II 9.7 1.04 0.38-2.82 0.875 Test for trend P = 0.897' To examine and control the potential effect of active smoking on cancer risk, the OR for lung cancer were analysed by smoking status. It was shown tfiat the OR in all exposure periods wereclose between~smoking pairs and non-smoking pairs; except in the period aged 1'S-22 years (Table 5). It is also noted'that the OR in all five age groups of exposure among non-smoking pairs were similar to those among all 114 pairs.. The lung cancer risk in relation to the level of house- hold PS was assessed by exposure period. The highly sig -cant 'dose-response' trends were observed for the three youngest age groups of exposure (Table 6). Table 7 shows the OR for lung cancer associated with the cumulative exposure tolhousehold tobacco smoke. No- ticeably, amongthe five age groups of exposure„only in the youngest is the OR statistically significant' (test for trend P < 0.001). DISCUSSION The effect': of PS on cancer, especially on female lung cancer, has been of increasing concern since the early 1980s. This is partly because, comparedJoimainstream smoke„sidestream smoke has three times as much benzol (a)pyrene, six times as much toluene, and more than 50 times as much dimethylnitrosamine;t" which may cause as many severe health problems as active smoking.15 227 However, how to measuresidestream smokeexposure and its effect on health remains a key issue. Several studies took the husband's smoking status as an estimate of PS for wives but this is far from accurate for the following reasons: (1) other family members, such as married relatives„parents and children may be important sources of sidestreami smoke; in some oriental countries such as China, this isparticularlytrue,(2)busbands may smoke outside the home, thus exposure does not occur, (3'):also„information about exposure before marriage is naavailable. It is reasonable to believethat if PS has an effect on lung cancer risk, the nature and extent, of t'~hae risk during childhood differs from! that during adult- hood. In this study, household exposure to tobacco smoke, from husbands and other family members, was taken as an estimate of PS, and collecte& for each resi- dence;,from childhood to adulthoodL As shown in Table 2, household exposure to other family members' smoke accounted for about one-third of total sidestream smoke during adulthood, which should be taken into account when estimating PS and its effecvon health. This study found that the OR for lung cancer associ- ated with.household PS varied by exposure period. This may be partly a result ofchanges in exposure with respect~ to frequency, intensity, and duration. It was indicated that cancer risk was highest for those exposed under the

22$ INTERNATIONALJOLIRN ALOF EP1DD EM1OLiOGY TABLE7' Ruk~of'Jemale~lungcancerjromPS~by exposure period~and'cumularin~ehouseholdrxposure~ro~,tobaccosrnoke;estinvued~b'}, ~ PSl'(gramstobaccolyearroj'exPosure)'in Harbin. China. Ezposure Casess Controls 95% period (age in years) PSI Nm N of No- total °Cd,of'~ total Odds ratio confidence interval P-~alue < 40 62 54.4 89 78.1 1.00 0-6 40- 24 21.1 14 12.3 2.46 1I 20-5.07 01015, 80- 28 24.6 11 9~7 3.65' f.74-7.76, <0:001 Test fbr trend P < 0.001' < 40 63' 55.3' 82 71.9 1.00~ 7-14 40- 31 27.2 23 20.2 1 75 0:94-3'.29 0.079 80 - 20 17.5 9 7.9 2.89 L 26--&63 0.012 Test for trend P=0.120 c40 68 59:7 76 66:7 1100 115-22 40 - 34 29':8 2& 24.6 1.36 0.7 5-2.47 0315 80- 12' 10:5 10 8;7' 1.34 0:55-3:29' 0.522 Test for trend P=0;350 < 40 77 67.5 81. 711.11 1.00 23-30: 40 - 30 263 28 24.6 1.13 0.62-2.06 0.699 80 - 7 6.1 5 4.4 1L47 0.45-j1.82 0,522 Test for trend P = 0.390 < 40: 48, 42:1 47' 41.2' 1.00 31-69 40- 34: 29.8 29 25'.4'. 1.15 0;61-2:17 0.67d 80- 32 28.1 38 33.3 0:83 0:4-5-1.53 0.543 Test for trendl age of 7 years ('OR 3.46, 95%. Cl: 1'.80-6.65) and increased twofold for those exposed to maternal smok- ing during childhood (OR 2.7, P = 0.003). There are several possible reasons for this: (7) exposure for children under the age of 7 may be longer an& more intense, because they usuaily remain in the home for longperiods and have close contact'with smokers, especially their mothers; (2) children are more susceptible to carcinogens than adults due to imperfect functioning of physicall metabolism„detoxifcation and immunity; (3) prenatal exposure, i.e. mother smoking during pregnancy. The children of parents who smoke have been shownto be especially'susceptible to respiratory problems that occur soon after exposure to environmental tobacco smoke.'e Correa et al:'7 in a large case-control study, found an increased lung cancer risk only for those exposed to. maternal smoking but not to paternal smoking. This finding complies with the result~ of our study. Findings from otherstudies support' the plausibility, of increased lung cancer risk from early life exposures to environmental tobacco smoke. An experimental study in animals have demonstrated transplacentalicarcinogene- sis with chemical compounds in! smoke.tg It has beem P=0:410 shown that compounds in tobacco~smoke can reach, the fetus via the placenta and may appear in breast milk.19'`0 EversonZ' also suggested'' that exposure to maternal smoking during'fetal'life could increase the possibility'of cancer inad'ulthood. Therefore„health effects, including cancer, are a major concern among children of smoking mothers as well as among mothers exposed to environ~ mental tobacco smoke. For PS, cotinine,22 nicotine23 and thiocyanate,24 the markers of PS, have been found after exposure to tobacco smoke. In additionl elevated blood! levels of carcinogens in passive smokers were also re- ported?5 Recently, an autopsy study?6'found significanti epithelial lesions among deceased non-smoking women who were married to smokers compared to those married to non-smokers. Findings from epidemiological studies''g•1D'17'27 have provided: further evidence for the effect of PS in early'life on cancer risk in adulthood. All of these studies found an increase of cancer risk in adulthood associated with early life exposure to parents' smoking, although the risk from exposure to paternal or maternal smoking was not of the samepattern. In a matched case-control study of 191 non-smoking pairs, Janerich et al.10 found househoU

PASSIVE SMOKINGAND THE RISK OF FEMALE LUNG CANCER 229 exposure to 25 or more smoker-years (which was calau~ lated by multiplying the number of years the subject lived in each residence by the number of smokers in that residence)duringchildhood and adolescencedoubled the risk of lung cancer (OR 2.07, 95% CI: 1.16-3:68), but no ;uch increase in risk was observed for adulthood expo- sure orlifetimeexposure. Recently; anothercase-control study6 also found an increased lung cancer risk for non- smoking women who reported 22'or more smoke-years of househol& exposure during childhood and adoles- cence (OR 2.4, 95% CI: 1.1-5.4), The findings from these two studies correspond well with our results. With respect to the effect of cumulative exposure of PS on lung cancer Dalager et a1:3 using pack-years of exposure as the measure, found a significant trend in the risk of lung cancer associated with the cumulative pack- years of exposure. In this stvdy; we attempted toestimate the cumulative effect of PS by using a passive smoking index (PSI). Using this index, we observed a significant increase in the OR with PSI for those exposed under the age, '7 (< 0:001). This finding is comparable to that of Dalager„and~ it agaiiiindicates the importance ofearly life exposure for the risk of lung cancer. It was suggested in, this study that, if the relationship between PS and lung cancer is studied only for adult exposure or over lifetime, the health effects of exposure during childhood may be masked. As a result, the whole or real effects of PS, especially that of early life exposure, may be underestimated and may have even escaped our attention in a; previous study.28 Failure to find elevate& lung cancer~ risk during adulthood and lifetime exposure from Janerich et aJ:'s report 10 provides further evidence for this view. The frequency and intensity of exposure to other people's tobacco smoke may change with~ age. For Americans, peak exposure occurs during a person's twenties! Besides, susceptibility to carcinogens during ahildhood may differ from~that in adulthood': Therefore„ coll( '-)n of exposure information from early life to adultnood and analysis of data by age of exposure is warranted. It should be noted that the finding ofa'dose-response' tendency in the three youngest exposure groups (P < ')_000 in our study may be questioned, because some of the'passive' smokers were also activesmokers, Matching by smoking status over lifetime betweemthe cases and controls ean, to some extent„obviate this criticism. Ih~a matched case-control study, Sandler et a1:9 investigated whethercancer risk in adult life is related to transplacen- tal or childhood exposure to their parents' cigarette <moke. Data were collected and analysed from the indi- viduals who lived withboth~ natural parents for all or most of the first 10 years of life. Of all the subjects ,tudied, there were 45% of the cases and~ 47% of thee controls who never smoked. One of their interestingg findings was that the relative risk estimates for smoking- related cancers (cancer of the oral'cavity and pharynx, oesophagus„pancreas; respiratory and intrathoracic or- gans, urinary tract and cervix) in relation to mother's or father's smoking were similar for smokers and non- smokers (0.8 versus 0:8 and 1.5 versus 1.7 respectively). In another case-control study,2' it was observed that overall cancer risk increased significantly withincreasing numbers of household exposure to tobacco smoke and the trend was similar between smokers and non-smokers. In our study, we included 59 cancer patients who ever smoked. Iman attempt to limit the effect of smoking on the lung cancer risk, individual matching by smoking status over lifetime (current- or ex-smoker versus never smoked over lifetime) was adopted. So, the cases and controls werecomparable with respect to smoking status over lifetime (ever smoked'or never smoked). For those 59 smoking pairs, there were no significant differences between cases and controls for the age smoking began (19.1 versus 20.7'years)~and for years of smoking (32.6 versus 31.1 years). However, the number smoked each day for the cases (mean 13.4 gramlday, SD 7.72) was slightly higher (P = 0!013) than that for the controls (mean 110.2, SD 6.41). This difference might magnify the observed OR in this study but it is unlikely to distort the true relationship between PS„especially PS imearly life, and the risk of female lung cancer~ observed in this study. Similarityof the OR infour out of five exposure periods amongst smokers and non-smokeris suggests no interac- tion, between passive and active smoking, Of note, the OR iniall fiveageperiodsamongnon~smokingpairs were similar to those among all 114 pairsstudied. These results again suggest that active smoking is less likely to have a strong effect, ifany, on the observed'association between lung cancer an& household exposure to tobacco smoke in the present study. Information bias is "inherited' in case-control studies, At present~ information on past exposure to environmen- tal tobacco smoke is often obtained by subject recalli through int'erview. It hasbeenfound thatpatients usually attempt to explainitheir illness.'91n this study, interview- ers were not blinded'to the subject's case/control status although they did not know the underlyinghypothesis of the study. In an attempt to limit information bias, we usedi structured questionnaire and standardized inter- view, techniques„which are often helpful in minimizingN both recall and interviewer bias. Also„it appeared thatQ collecting information on PS by each residence may help jv subjects to recall past exposure to environmental smoke„{~ especially for those 'yes or no' qgestions. Coultas et al'10AQb reported thatinformation on whether there was exposu re+4 to parental smoking duringchildhood was more reliableW cc GO

230 I N'pFR'NATIONAL JOURNAL OF EPIDEMIOLOGY than thati on how much exposure there wasin the same period. In this study, however, we were not able to validate the histories of past exposure and'' we did not have information onifetal exposure to tobacco smoke. Also, our data were limited by the small numbers of specific histological types of lung cancer which limited sonie interpretation of the results. It should aiso be noted that the latency period' between household exposure to cigarette smoke and diagnosis oflungcancer may be veryy long. If so; the examination of the household PS'frorn ages 311 to 69 in our study, thereby excluding cases over 69 years, might underline the effect of PS for those exposed only during that age per'ibd.. Nevertheless, the findings from this study provide further support to the observation that PS mayincrease the risk of subsequent lung cancer„especially PS during childhood. It is also suggested that assessment of PS' should be done by different exposure period's. REFER'ENCES. t Trichopoulos D, KalandidiiA, Sparros L and MacMahon B. Lung, cancer and passive smoking. InrJ Cancer 198!1; 27: f141_4 8. ' Hirayama T. Cancer mortality in nonsmoking women with smoking, husbands based on a large-scalecohort study in Japan. PrevMed 1984;13:680-90! 3 Dalager N A, Pickle L W, Mason Ti 1, Correa P, Fontham,E er al.. Thc relation of passive smoking to lung cancer. Cancer Res 1986; 46: 4808-II. ° Fontham E T, Correa P, Wu-Williams A, er al. Lung cancer in non- smoking women: A multicenter case-control study: Cancer Epidemiol Biomarkers Prev 1991; 1: 35-43. 5Pershagen G, Hrubec Z and Svensson C- Passive smoking and lung cancer irt Swedish women. .-1m J, Epidenriol 11987; 125t 17-24. ° Stockwell HI G, Goldman A L. Lyman G' H' et' al' Environmental tobaeco:smoke and lung cancer risk in nonsmoking,women. J Natl Cancer Insr 1992; 84: 1417-22. t Sandler D'P, Everson R B; Wilcox A J, Browder J P. Cancer risk in adulthood'from early life exposure to parents' smoking, ,4m J Public Health 1985; 75: 487-91. e Janerich D T. Thompson W D~ Varela L R er al. Lung cancer and exposure to tobaeco smoke in the household! N EnglJ Med 1990; 323:632-36. 9Chilmonczyk B A, Knight G J, Palomaki G E, Pulkkinen A J, Williams, and Haddow'J E. Environmental tobacco exposure during in- fancy. Am J Public Health 1990; 80i 1205--08. 10 Jarvis M J, Strachan DT, Feycrabend C. Determinants of passive smoking in children in Edinburgh4 Scotland. Am J'Public Health 1992; 82: 1225-29. rt Friedman G D, Petitti D B, Bawol R D. Prevalence and correlates of passive smoking:. AmJ PublicHealrh 1983; 73: 401-05:. r2 Pron G E, Burch 1 D, Howe G R', Miller A B. The reliability of'passive smoking histories repo!raed in a case-control study of lung cancer. Am J Epidemiol 1988; 127: 267: 73. 13 Qian Y P. Advances in €prdemiology, Be jing Press of'rhe Peo}; e s Ilealrh 1985; 3: 197-207. ~ US Department of Health and Human Services The I7ealth Co>•,<e- quences of Smoking - Cancer A repart of the Surgeon Genera! D}}HSpublication no. (PHS)'8;?-50179;,1982. 1sSpitzer W'O, Lawrence V; Dales R eral~ Cinksbctween passivesmoking and discasc A best-evidence synthesis- Clin Inwsr Afed 1990; 13: 17--02. 16 US'Department oflHealth and Human.Services. The Health Corse- quencesoflhvofuntarySmokingr A Reportof theSurgeon General' Washington, D C: Goverrtment' Printing Office, 1986 DHHS publication no. (CDC)S7-839%',)1 rCorrea P, Pickle L W, Fonth'am E, Lin Y, Haenszei W: Passive smokine andlung cancer. Lancer 1983; ii: 295-97. 18 Drew R'T, Boorman G A, Haseman J K, McConnell E E, Busey V1' M, M oorcJA. The effect of age and'exposure d uration on canc~eri induction by aAnown carcinogeniin rats. mice, and hamste-,s. Toxic Appl Pharmacol 1983; 68: 120-30, 19 Luck M and Nau H. Nicotine and cotinine concentrations in serum and urine of infants exposed via passive smoking or milk from smoking mothers: J'Pediatr 1985; 107: 81 C-820. 10Woodlvard A„Grgunnovich N and Ryan P Breast feeding andl smoking hygiene: major influences on cotinine in unne of' smoker's infants. l Ep, idemiol ConrmunitJ Health 1986; 401 309= 1'5. 21 Everson R B. Individuals transplacentally exposed tornaternal smok- ing maybe at increased cancerrisk in adultlife Lancer 1980: u:. 123-27. ` Wald N'J, Borcham J, Bailey A, Ritchie C, Haddow J E, Knight G Urinary,eotinine as marker of'breathing other people's tobacco, smoke. Laneer 1984, i: 230-31. 27 Greenberg R A„Haley N'), Etzel R A, Loda F A. Measuring the exposure of infants toYobacco smokc: Tdicotine and~.cotinine in urine and saliva. N'Engl J Med 1984; 310: 1075-78. '° Bottoms S F; Kuhoert,B'R', Kuhnert P M; Reese A L. Maternal passive smoking and! fetall serumi thiocyanate levels. Am J' Obsterr Gyneco/,1982;,144:787-91. 's Maclure M„Kataz R B A, Bryant:M S, Skipper P L, Tannenbaum,S R.,Elevated blood levels of caroinogens in passive smokers. Ain J'Public Health 1989; 79: 1'381-84. '-6 Trichopoulos D, Mollo F, Tomatis L er al. Active and passive smoking and~ pathologicallindicators of lung cancer risk in an autopsy . study. JAMA 1992; 268: 1697-701. '7 'Sandler D P; Wilcox A), Everson R' B! Cumulative effects of lifetime passive smoking on cancer risk. Lnncer 1985; i: 312-1 W. 'e Wang F L,Dai X D; Liu N, Liu C'B. Analysis ofrisk factors for female lung,adenocarcinoma in Harbin; Indoorair pollution. China J Prev Med 1989; 23: 270=72. .9 Sarewell P E. Retrospective studiesc a review for the clinician. Ar.n Intern Med 1974; 81: 381-86 70 Coultas D BiReake G T and Samet J M. Questionnaire assessment of lifetime and recent exposure to environmental tobacco smoke. Am J Epidemiol 1989; 130: 338-47. (Revised version received October 1993)