Tobacco Documents Online

73' THE NE%~ ENGLV\D JOLRV.,,LAF M EDIICIN E Irpi -). ,,,1. LUNG CANCER AND EXPOSURE TO TOBACCO SMOKE IN THE HOUSEHOLD Dt.'tGHT T. JANERIcH. D D.S.. \I.P H.. \V. Dot:GL.ks THOMPSON, PH.D.. Ltrts R. V.+RELA. ~f.D'.. PH D.,*PETER GREENWALD. M.D.. DR.P.H,., SHERRY CHOR05T, tI.S:. CATHY Tl.'cct. B.S.., ML'HAUMAD B. Z.A>l:#N. M.D.. \Il'ROV R. MELAMED. M.DD., %IAL'REEV I+aELY, R.N.. AND MART1V F., McIti.NE.+LLY. M.D.. Abstract Background. The relation between passive smoking and lung cancer is of great public health impor- tance. Some previous studies have suggested that expo- sure to environmental tobacco smoke inithe household can cause lurx} cancer, but others have found~ no effect. Smoking by the spouse has been the most commonly used measure of thls exposure. Methods. In order to determine whether lung'cancer is associated with exposure to tobacco smoke wtthtn, the household, we conducted a populi3tion-based case-con- trol study of 191 patients with histologically confirmed pn- mary lung cancer who had never smoked and an equal number of persons without lung cancer who had never smoked. L,fetime residential histones including informa- tion on exposure to environmental tobacco smoke were compiled and analyzed. Exposure was measured in terms of -smoker-years,"'determrned'by multiplying the number of years in each residence by the number of smokers in the household. Resufts. Household exposure to 25 or more smoker- years dunng childhood and'adofescence doubled the risk of lung cancer (odds ratio. 2.07;; 95 percent confidence interval, t.16 to 3.68). Approximately 15 percent of the control subjects who had'never smoked reported this level of exposure. Household exposure of less than 25 smoker- years dunng chikihood''and adolescence did not increase the nsk of lung cancer. Exposure to a spouse's smoking, which constituted less than one third of total household' exposure on average, was not associated with an in- crease in risk. Conctusions. The possibility of recall bias and other methodologic problems may influence the results ofcase-l corttrol studies of environmental tobacco smoke. Fbrtetfte= less, ourYindings regarding exposure during early life sug- gest that approximately 17 percent of lung cancers among nonsmokers can be attributed to high levels of, exposure to cigarette smoke during childhood and adolescence. (N Engl J' Med 1990; 323:fi32-6.). THE 1972 Surgeon Gencralls report dealt with the health consequences of passive smoking or envi. ronmental tobacco smoke for the first time.' In 1986 the entire reportwas devoted to the issuc: it concluded that "inaoluntarv smoking is a cause of disease includ- ing lung cancer in healthy non-smokers."' More than a d'ozen epidemiologic studies have assessed the rel'a- tion betwcen passive smoking and lung cancer."'' The findings have ranged, from no detectable increase in risk" " to a moderate (about twofold), statistically sig- nificant increase." MMost studies have found only small elevations in nsk„which are frequently not sta- tisticallv significant.3'°•9 In a meta-analysis of all the available studies in 1986, %tiald et al. found a slightly increased risk of lung cancer associated with environ- mental tobacco smoke.1° li,Ve undertook the current study in an attempt to clarifv further the role of passive smoking in causing lung cancer. In this report we discuss exposure to to- bacco smoke in the household as a possible cause of lung cancer among nonsmokers. Mtraoos We conducted a population-based. individuahv matched case- control study in New York State from 1982 to 1985. The cases were drawn from seven Standard Metropolitan Staustical Areas From the Departmem of Epidemio"r and Public Heilth. Yak Universtry Schoot,of Medicine, New H.ren. Conn. tD T.1.1: the Unrvers ry of S.outtiem Matne. Ponland Iw D T.): ttie Populuwn Cowrcil of Mexico. Mexico City iL.v R.r tbe Nanonal Cancer lnsutute. wnAington. D.C. IP G.I: the Ne. Yort State Dep.runem of Health. Atb.ay (s.C.. C.T. ): the Skten-Kearnng Memonal Institute. New Yort IM 8 Z.. M.R.M.1: ard Albany Medical Colk:e, Albany. N. Y- t M K.. M. F. M. t: Address tepnnt requeus to Dr. lanench at dvr Depnn- mem ot Epidemtolofy and Publx Health. Yak Umventry School of Medicine, 60 Coilere St. tLEPH 1051. New H.ven. CT 06510. Supported in part by pants tP01 CA 42101 and ROI CA 3206bI Gvm th Nattonal i Iesututes of Health. •Ikceated. iBuffalo. Rochester. Svracuse. Uiica-Rome. Albanv-5chenectad.- Trov, Binghamton. and greater New York. excluding the fi.e bor- oughs of New York Cin t. This geographtc area compnses 23 coun- ties. with approximatelv 125 diagnostic or treatment fact6fin, and a population base of nearl4 10 mdlion people. A specul'sl stem for the rapid ascertainment of cases ofllung cancer was establtshed in these 125 facilities so thatpatients could'be identified and enrolled as soon after diagnosis as possible. All new cases of lung cancer diagnosed clinicallv, histoiogualli; or bottil were regularl% identified at the participating hospitals. The `ew York State Cancer Re.,tstn was checked routinely toidenufo anv cases that might ha%e been missed by the hospitaUbased reporting svstem. Information on smoking,was initially obtained itom the patients' medical records. .ill the ca.se patients reponed' as havtng never smoked or as former smokers or whose smoking htstorv was un- known were contacted by telephone. and their smoking statuswas confirmed. To be included as a-case" in the studv, a patient had to reside in the 23-countv area. be between 20 and 80 vearr of age, never have smoked more than I00 cigarettes (nonsmokcrs)or have smoked at some time but not have smoke&monr than 100 cigarettes in the 10 vears before diagnosis lfontter smok'enl. and have been given a diagnosis of pnmarv lung cancer between July I. 1982. and December 31. 1984.,that was confirmed on reexamination of the patholoSical specimens and clinical records. Slides or blocks of' tis- sue were available for all but five of the case patients. All matenals were reviewed bv investigators who were blinded with,respect to the patient's initial diagnosis. smoking historv. and'other nsk factors. Interviews were conducted with 76 percent of the eligible pauents or their closest avatlable relatives or friends t surrogatnl. Control subjFcts were individually matched to the patients and were selected by screening the files oC the New York State Depart- ment of Motor Vehicles. This source of controls was considered appropriate since itwas population-based and provided most of the infotmation neceasarv to perform the matching. A list of potential control subjects for each case patient was selected on the basis of age (within five veanl, sex, and countv of residence. Potential con- trol subjects were contacted by telephone. The first eligiblt subject who was found to match the case patient in terms of smoking histo ry(nonsmoker or former smoker) and who agreed to participate was enrolled in the study.Art additional matching vanable considered att the time of data collection was the type of interview - i.e.. direct interview with the patient or control subject versus interview with • surrogate respondent: When a surrogate cax patient had to be interviewed„we also interviewed a surrogate for his or her matched 2023382'39t7

%ol 3:3 \o 10' LL.NG CA`CER_AND HOISEHOLD TOBACCO SMOKE - JANER'ICH ET fill, control. even,when the control subject was available and willing too be Inter.aewed. Further information on the methods used in the stud~ is a% ailable elsewhere.'s Data were collected for 439 case-tontrol pairs. Of these. 242 patn ,.ere former smokers and 197 pairs had never smoked. Separating the restdual!dfects ofdirect smoking from those of'passis e smoking among former smokers in.olYes more complex analvtsc and tnter- pretanonaliissues than does an examination of the effects of passive smokln¢ in ~hose .•ho ha%e ne%ersmoked. This report is therefore limited to:prrsons ~, ho nr.ersmok'ed. Six of the 197 patn who had ne%er smoked ..ere mismatched tniterms of the type of interview direct %s. surro¢ate-and hace therefore been excluded. Thus' the anal%ses reported here were based on 191 matched case-control patrs A total of 129 pairs wrre rntentewed direcdv, and surrogates were interstewed for 62. AII inl'ormaoon was collected during a face,to-face interview with use of a precoded questionnaire. Case patients and controfsubjects were mter*tewed in exacdv the same fashion, and except for items concerning the clinical' aspects of' the current medical condition. both groups answered the same questions.. Information about smoking in the househol&was collected xpa- ratel. for each residence in which the subject had lived for one vear or more. up to a maximum of 12 residences. The number of "smok- er-.ean" of exposure was calculyted~bv mulupli tng the number of .ears the sublect lived in each residence by the number of smokers nneludin¢ the spousel in that residence. The producta for all rest- dences were summed. Smoking bv the spouse was also recorded separately from that by other householtl'memben in a subsequent section of the questson- natre. The informanon conststed of the number of'vean the spouse had smoked while living with the case patient or control subject and the number of cigarettes smoked per dav,. Smoker-vean of exposure from the spouse's smoking were calculated in the same manner u for the entire household. Pack-vean of exposure from the spouse wrre calculated by multiph ing the number of' packs smoked per day b% the number of vean thatahe spouse smoked while living with the subject. lf the subjecrhad been mamed to more than one smoker, then the numbers of smoker-.ean and pack-years of exposure for all spouses were summed. The questionnaire also included sections on exposure to environ- mental!tobacco smoke in the workplace and in social settinp out- side the home. The format for these questions differed from that used to collect data on exposure in the household. The summary results of this anal.xis are presented here; detailed finding are acatlablt elsewhere."' Statistical techniques appropriate for the analysirof individually matched case-control studies were used.' For clarity of presenta- tion. percentages were tabulated for case patienu and control sub- jects sepantelv. rather than for matched pairs. However, odds ra- nos were computed on the basis of the matched pairs. The conditional, logistic-regresaion tnodel was used in the muluvariate analvses." Comparisons of the effects of exposures that occurred during different periods of the subjects' lives were based on evalua- tion of differences in the magnitude of appropriate logisuc-regres- sion coefTscients. For statistical testins of thex di6crcnces. we used the vanance-covarfance matrix fivm the ksgisuc-re6ression anal}ses. RrsUn.rs Smoking by spouses contributed a large propor- tion of lifetime exposure to environmental tobacco smoke but was not the chief source of exposure. Table 1 shows the amount of exposure to environmental tobacco smoke (expressed in smoker-years) during childhood and adolescence, during adulthood, and from the spouse for the 191 control subjects who had never smoked. There were only small differences be- tween memand women. The spouse contributed about 30' percent of the lifezimt smoker-yean of exposure; the correlation coefficients for exposure from the spouse and lifetime exposure were 0.37' for men and 63} Table t. Distnbution of Smoker-Years of Exposure to Envlron- mental Tobaoco in the Household. • G.nr.o.r or. Esrowu NYM w,,w,t. l.ifettmc srroker-yean. 46 6xS3.7 trtcart :SD Smoker-years dunnt childhood and adokscencet 52 7_a2.9 Me.n cSD 154:206 I6Im162 Percent of lifetime exposure 33:J 30~6 Cortetauon with hfetttne exptxute 0 92 Smoker-yean from~spouse 0 ~61. Mean _ SD 13 0_ 17 0 16 2= 16 7 Percentof,tifenme ecpowre .e.0 30 7 Corre.lauon wieh lifeume eaposure. 0 37 0 Sr Smoter-yean dunnrg adultlsoo4ftotn sources other than spouse Meae, _3D IA1_3I0 ?0S_=99 Pcrcent of lifetime expotue 38 9 36 9 coreelation wttli lifeutne exposure 0 91! 0!3 -eYed al l7.nNle Md I66.rtnWe ealqotl wApru MOo htlR.er 1mWe0 ~ nWf Ioo . np,are..Infamsbo. oa.unoter-reand eafowe fniiri 111e cpmre .% ta e7.mab nC Iep fem.le cwnd sWtensmote.yean .e,e cNcvle.Q b7.-OnOtyin{ tM mmlier cf.rerf Ne weMct Ind uo a nedeeeT tryy nr suml.r of unet.n u tlr ~o.rdnW 1tae tsr 21 yen of rs. 0!51 for womenl Exposure during childhood and ado- lescence (<21 vears of' age) contributed a similar per- centage of the lifetime smoker-years but correlated more closely with lifetime exposure (correlktion coeEF- cient, 0:92 for men and 0.61 for women). The average lifetime exposure was 46.6 smoker-years for men and 52.7 smoker-years for women. During adulthood, household exposure from sources other than the spouse was somewhat greater than from the spouse. Table 2 shows the odds ratios for d'eveloping lung cancer in relation to the degree of exposure to tobacco smoke in the household for the l91 nonsmoking case- control pairs. The data are stratified by la'vels of expo- sure (measured in smoker-years) and by the periods of life when the exposure occurred. Exposures during childhood; and adolescence were defined as exposures that occurred when the subjects were less than 21, years of age. Exposures during adulthood include all household exposures from 21 years of age to the time of diagnosis. Although the odds ratios for the higher exposure categories are somewhat higher than those for the lower categories, no clear dose-response rely- uon is evident, and most of the 95 percent confidence intervals include 1.0. For exposures in childhood and adolescence, the highes`t l'evel' of exposure is associated with the greatest elevation in risk, and the 95 percent confidence interval excludes the null value (odds ratio, 2.07; 95 percent confidence interval, 1.16 to 3.68). For the 129 case-control pairs who were interviewed di- rectly, the odds ratio for persons with 25 or' more smoker-years of exposure in childhoodd andiadoles- ,ence was 2.31 (95 percent confidence interval, 1.1& ,~'rto 4.61). With smoker-years during childhood and adoles- cence and smoker-years during adulthood treated as continuous variables and included siatultaneously in a logistic model, each increment of five smoker-years of exposure during childhood and adolescence was found, to increase the risk of lung cancer by 6.5 percent (95 percent confidence interval, 0:l to 13.2). On the other hand, each additional five smoker-years of exposure

I 634 THE NEW EVCLA.VDJOL'RtVAL OF 14EDiCI`E Sepf o, 1490 Table 2. Relaton,ot Smok.r-vears of Exposure to Ernnronmentai Tobaoco Smoke to dt» Risk of Lung Cancer alnonq-Persons WAo n7ever Smo+(eC Mon tnan 100 Ciqarettes.I •o 13. cA,. S.waa.Yaw PanmMn cc.r.ou Ooa4.ncM% Qri' ehilmood ,detexeixt• 0 s'/ :9a1 6e 05 6) 1,:. 82 1,2 9) 9. (.9 2) I 09 (0 64- r 77) s2! !2 t27:9 1 tJ 2) : 2 07 (I 16-3 641 itlu4tBootl o +•230) 39120 4) I-:• 37 I')9 41 +a i2.1 Ii 066 i0.34-1,21) s-49 +61u1) 50 126 2) 0 11, to.s-t .s). w--, 36(11 .9) 32 (16 e) l 1 0010 .32-1.931 2.,s ss I Ir 7) .:2 (11 s); 1 11 to 58-2 20) L,ftnmc 0 32(16et 33(173) 1-:s :0(I03) 27((.I) 078 (0.36-167) J-+9 37 (ii.3) 46 i2s 11 0,10 f0 0-1' 301. 50-.'• st i23 0) 40124191 1 19 10 63-2 271 •s_99 33 (17,31 Z1v(lu 0)1 1 90 (0 e3-3 90) b,oo :7 (14 1) :4I Iz.6) 11 u lo.se-2.2s1 w 191 m.cL.d c.rsoard {w C7 i0no.. ca~hOto ut~.r Ori noor am ~Ip.n fa.lpeno..~m m. nibwt.qic~0./ qM- d ~A t t~e ~r r.a~1 a.„ot.r r.n ~ 60~ or osy. vt,t.+w. ee " M..n Wr wo. •t.at st.2t.y..n at.pf during,adulthood were estimated to have virtually no effect on risk (95 percent confidence interval. -33 to +2.8 percent). The difference in the magnitude of the effect between exposure during childhood and adoles- cence and exposure during adulthood did not achieve statisticali significance (P = 0.12)I On the basis of the distribution of exposure levels during childhood' and adolescence among the control subjects and the magnitude of the effect of early exposure, we estimate that approximately 17 percent of all lung cancers in nonsmokers can be attributed to exposure to paa-," sive smoke in the household during childhood and adolescence. On the basis of the odds ratios for the 129 case-control pairs who were interviewed di'- rectly, approximately 19 percent of lung cancer in nonsmokers appears to be attributable to exposure to environmental agarette smoke in childhood and adolescence. Since smoking by the spouse has been the most commonly reported measure of exposure to environ- mental tobacco smoke in previous studies, we exatn- ined exposure from the spouse separately, although exposure to environmental tobacco smoke from, the spouse is also included in the results shown in Table 2. The odds ratios for exposure frequently differed ac- cording to the type of interview, especially for the data on exposure to a spouse's smoking. Table 3 there- fore shows the results of the analyses of exposure to environmental tobacco smoke from the spouse separately for subjects interviewed directly and' tbose for whom surrogates were interviewed. The odds ratio for the development of lung cancer for those who ever had' a spouse who smoked, as compared with those who did not, was 0.93 (95 percent confidence interval, 0.55 to 1-57) for those interviewed directly. In terms of smoker-years of exposure to the spouse's smoke. the results show little effect, with an,od'ds ratio of 1.07 for 25 or more smoker-vears of exposure t95 percent confidence interval, 0.39 to 1.97). Estimates based on pack-years of exposure from the spouse were similar to those based on smoker-years. For both measures, there was little evidence of a trend accord- ing to amount of exposure among those who were exposed. All analyses were repeated for only the case-control pairs for whom we had complete and internally consu- tent data for all residences and marriages. Anv pair was dropped from these analyses if data were incom- plete or missing for either the case patient or the con- trol subject, leaving 113 pairs of nonsmokers. Our purpose was to ensure that our conclusions were not dependent on the particular methods we adopted to handle inconsistencies or missing items in, the data set. The findings were similar to those for the entire group of 191 pairs. The odds ratio for exposure to 25 or more smoker-years in childhood and adolescence was 2.59 (95 percent confidence interval, 1.22 to 3.49); Exposure in the workplace was measured by record= ing the number of smokers who worked with each study subject during his or her lifetime and the amount of time the subjects spent working withh these smokers. These exposures were compared for case patients and control subjects. Estimating the odds ratio as a continuous variable for an equivalent differential of 150 person-years of exposure gave an odds ratio of 0.91 (95 percent confidence intervaL 0.80 to 1.04), indicating no evidence of an adverse effect of environmental tobacco smoke in the work- place. Our assessment of smoking in social settings used an untested, semiquantitative index in which the case patient or control subject used a score of 0 through 12 to indicate his or her regular exposure to tobacco smoke in social settings during each decade of life. Cumulative lifetime reported scores ranged from nearly 0 to more than 70. The odds ratio for an increase of 20 in the cumulative score was 0.59 (95 percent confidence interval, 0.43 to 0.81). Our analysis of exposure in social settings with use of this index showed a statistically significant inverse as- sociation between environmental tobacco smoke and lung cancer. D11ct13ilON We found a statistically significant adverse effect of relatively high levels of exposure to environmental to. bacco smoke during the early decades of life (up to the age of 21)' For those wbo were expo.ed to 25 or more smoker-years during their first two decades of life, the risk of lung cancer doubled. This amount of eaposure is equivalent to living with more than one smoker throughout childhoai' and adoiesceace - a high but na uncommon level of exposure. An etpo.nr,r of this level was reported for approximately 15 percent of the control group. By coatrast, we found no adverse effect of exposure to environmental tobacco smoke during adulthood, induding espcsurc to a spome wbo 1

! %ol. 323 ', - o. IU LLNG CANCER AND HOLSEHOLD TOBACCO S.MOKE - JANERICH ET NL Table 3. Relation of Spouse's Smoking to the Risk of Lung Can- cer among Persons Who Never Smoked More than t00 Cga- rettes. According to Type of Intervlew.• ORLC7 ad1ls ..nn -97F C! i E.er had a spouse .Iw smoked'No - Ves 0 93 (0 53'-1 57)'. Smoker-vean of caposue from spouse SL t1OG,TI 0 u 10 1q=I 02s 0, - 1-.d o 7! lo Q-I !01 0 33 (0 f1 -1 03) ;023 I 07 10s9-1 97) 0 33 io 1:-0:95) Pack•ysars ol exposure from spo 0 ux - 1-2t 0 71 /0 37-1 37) 0 1e (0 ,04-0 621 25-49 098 10 47_2.03) 069 I018-260) ;1, 50 I 10 lo t7-2 56) 0 20 1o 03-1.221 • eueE ba 129.car~awef pues imerrrred dveNr.mt 79 Mu+ for +nom wemo.n +en ,nnenwred flm of Cm 191 pun +en e.cduOeE lac.uu for ons.mcmoer of tlr pur tlsve .r mi» snt mfotmnton sbws +Ixdier.Qn suErn luda sCwn.ln unok.A. Dr+oe tmoYec:. .c.n of espesun .en a.ud.Dkfor 129 ew~onvol IPun .,m duea insnK.s +nG S6 pum. .,0% su"ns ~mtni.rs D.u on p.cl-rean.of espown we s.ulaE/e for 1229un..'ds O- ~nsm.sa..nd31 pun..rth wnot.e ~nerrrraCi d~ rn+hOrnce ~naenH Odds nuos H sAO..n lor.rr per+on a1h Or c19wurI specJrOV comprW adsl peno..0 no espowre to a spoyr .#o unW.d smoked. Although problems of recall and other poten- tial biases may have influenced the results, our data suggest tharexposure in early life may be a limited but important contributor to the risk of lung cancer in nonsmokers. A previous study with a smal) number of subjects found little evidence of an elevated risk of lung cancer, among nonsmokers whose parents had smoked." Children of parents who smoke have been shown to be especially susceptible to respiratory prob- lems that', occur soon after exposure to environmental tobacco smoke.' ' This type of susceptibility might initiate changes that eventually lead to lung cancer when the exposed children become adults, but we know of no specific mechanism that would explain our findings. We found no adverse effects of exposure to tobacco smoke in the workplace, although we did noo have enough information about the level of exposure in the workplace to assess the precision of our measure- ments. The apparent protective effect of exposure in social settings is difficult to explain. During the course of this study, regulations in New York began to restrict smoking in the workplace and in social set- tings such as restaurants. We did not anticipate this development and cannot estimate how much the awareness of these new restrictions might have af- fected the responses of the study subjects or their surrogates. Evidence is clearly mounting that tobacco smoke inhaled passively by nonsmokers is potentially car- cinogenic. In a recent study, Maclure et al."' found elevated levels of carcinogens in the blood of passive smokers. Levels of hemoglobin adducts of 4-aminobi- phenyl and adducts of 3-aminobiphenyl were signifi- cantly elevated in subjects with con6rmed exposure. The validity of this finding was supported by addi- tional evidence that showed a sharp decline in the levels of adducts among smokers who quit." At present, information on paso exposure to en%i- ronmental tobacco smoke can be obtained only by interview. The available biologic markers, such as co• tinine, cannot be used to confirm exposure that oc- curr~ed years or decades earlier. The use of interviews to obtain a lifetime historv of exposure to passive smoking requires that the questionnaire be structured and the interview techniques be standardize&so that all subjects are interviewed in the same wav. We took steps to ensure such standardization. Two recenn re- ports may lead to improved ways to measure lifetime exposure to environmental tobacco smoke by means of interviews.r"' In one of these studies, which at- tempted to evaluate the reliability of interview data by repeat interviews, information on exposure during childhood was found to be very reli'able." It was necessary to use surrogate respondents for about one third of the interviews, usually because the patients were too ill to be interviewed. To minimize potential bias, surrogates were also interviewed fbr the matched control subjects„and separate estimates were calculated'for respondents interviewed directly and surrogate respondents. We used equal care in all types of interviews and in all subject areas covered in the interviews; however, the data we obtained in inter- views with surrogates still differed somewhat from those obtained in direct interviews. Inaccurate report- ing of exposure tends to bias odds ratios toward' the null value unless a systematic bias is present. Data from surrogate respondents are likelv to introduce random error because of the surrogate"s Iack of de- tailed' knowledge of the subject's exposure. On the other hand„ it is possible that the surrogates for pa- tients with lung cancer might tend to underreport the exposure contributed by their own smoking to a great- er extent than surrogates for control subjects. Such a difference could mask an actual increase in risk or reverse the direction of the association. The findings shown in Table 3 indicate that the use of data from surrogates may have led to an underestimation of the effect of exposure from the spouse. :Vthough our re- sulu for exposure due to smoking by the spouse differ from those of earlier studies,'" our 6ndings regarding other types of household'exposure support the conclu- sion that exposure to environmental tobacco smoke can cause lung cancer. Akiba et al.,' Dalager et al.,' and Garfinkel' have reported elevations in risk of 30 percent, 50 per- cent, and 10 percent4 respectively, associated with ex- posure to a spouse's smoking; none of these increases were statistically significanr. With the exception of Chan et al." and Koo et al'.2s in Hong Kong, these and most other investigators have reported point estimates that suggest an increased' risk for those exposed. The duration of exposure, as measured by the number of years the spouse smoked while living with the sub- jcct, did not have a statistically significant effect in our data. Two studies that used the same measure

636 THE NEW ENG WdD JOURNAL OF MEDICINE of exposure also failed to exclude the null value.",9 Garfinkel et al.,"using a different measure for duration of exposure (husband's smoking in the last 5 and 25 vears); found one significant association among the large number examined. Exposure due to smoking b.• the spouse. expressed in terms of pack-years while the spouse was living with the subject„ was found not to be significantlv associated with lung can- cer. Using a comparable measure of exposure, Tricho• poulos et aP reported relatively large increases in risk (greater than twofold). Perhaps our data do not show that smoking by the spouse increased the risk by itself beuuse smoking by the spouse made up only about one third of the subjecu' lifetime exposure to environmental tobacco smoke? It is also possible that physical circumstances and differences in study areas. the size of residences, ventilation, and other important physical aspects of the living conditions, as well as social habits that affect exposure within the farrtilv; will' nee& to be measured and analvzed before the differences in findings among the studies can be reconciled. The evidence we report lends further support to the observacion that' passive smoking may increase the risk of subsequent lung cancer, and it suggests that it may be particularly important to protect children and adolescents from this environmental hazard. 44'e are indebted to Andreaa Nicolaou for his auistance with the computer programming used in our analvses. RR.FC]RzA/f:JLS I. DeQaratrnt of Healm. Educanon. .nd Welf.re The Aeanh coeneqtrnees of wmtmg a nrport of the Sursem Graval 1972. wasAmpon. D C. Gov. ertunem Pnnnng Office. 1972:121-33. (DHEW pu6fkauon,no (HSM)72- 7916 ) 2. Dep.ratent of HeaMt aed Human Sernaea. TTe hraltb wroaquesces of in.olunuur .rnotLtt: a reyon of tAe SurBeoa Gtnerfl. W.utunjwn.,D C::, Govemmeat Pn>mng Offlcs. 1986. (PubUc.noa no. DHHS (CDCl i E7- e39e ): 3. G.r}inhel L Time treods in 1ung cancer teudiry .rnong narmotaf wd a eae oa pnnve fmokm{. J Natl Caeoer lst 19BIL 66:1061-6. Sept. 6, 1990 a: Hiny.m. T. Canea nwrWxy m nor.eooring romen ~da rmoitpnf hua. E.eda tased on a Irse-acale cohart andy, ia l.p.n. Ptrr Mad 196t'. 13 66t1a 901 S Tnctroqoulot D. Kal.ndidi A.,Spvtoa L Lung cancernd paeave emoting cwiclauion ofGeeet nudyLaicet 1963: 2:677.{0. 6. Correa P. Pxkle LW: FomAua E. L+n Y: H.enucl W: P.suve amo/ung and lune cancer. Lancet 19l3: 2:395-7. Garhnkel L. Auerbwdi O. louWrrt,L. Iawluntrry, amokin{ rd lung cancer. a ca.ermmol .+dy. ) NrJ Caocer Irot 1915. 75 463-9 AkJb. S. Kro H. Blot WJ. Pwrve smdcin{ and lung cancer among lap.- nese .omee. Cancef. Res 1966: 464d04-7:. DUager NA. Pichk LW. Masan Tl. et at' 71K nelaoon o( psuive unajung to.lung canuer- Caneer Ru 1986:.46-al0l-I I. 7. ~. 9. lo., KaEa GC. Wyndcr EL. Lung cancer m noe-amters Cancer 196r. 53:121,a-2r. 1 I. Sandkr DP. Wikox Al. Erenan RB. Cumulaove effecn of lifeume pa.e ve smobnj on cancer.nak. Laocet 19es: 1!312-5. 12. Sudler DP. E•enon RB. W ikoa AJ. BnTrOer 1P. CaeQr ruk m edi+lmood from early life eapowre o pretw wnakung Am J Public Hka1th 1953. 79;4l7.92. 13: Sasdke DP. Ereraat RB. Wikox Al;, Paui.ti urcktnt in adultMod ard cancer nak. Am 1 Epidemat 1915: 121:37-48. I4'. PersAagen G. HfuEx 2. Svenaean C. Paaire vnoting and luty caeca in SreduA romen. Am l Eqedemat 1967. 125 17.21.. IS. Koo LC. Hb 1H. Saw D. Ho CY Measuaement of pawvtimokoil rd ewrtres of lun{ cancer nak anwng noe-emotiag CTiineac females. lnt J Gecer 1917; 39:162-9. 16. Hurnbk CG: Samet 1M. Padt.l DR Mrnaye so a amot'n and lung canrer nel. Am 1 P1ibIK HealtA 1997. 77.39A-602. 17: Cb.n WC. Fvrif SC. L:wq caeaer in ronumoLen in Hong Kong In: Grw+d- nurr E. Ckmmeaen 1. Muir C. ed1 GaopapRical p.rMlocy, ue wca eyidemiolop. New Yort: Gauar Fidmr Vertaa. 1962:199.202. II. Wild N1. N.ncA.haliK. T1wmPson SG. Cuckk HS Don Eeeaming odier peopie'e w6.eco emokk cauae IunB cancer' BMJ 1966: 293.1217.22- 19. Vaecla LA Annsmcs of drc u.ocutioa betreee p.uire smolung and hmf cmr. IPE.D. disansaoat. New Haven. Com.: Yak Univenity. 1997.) 20. , Bteslb. NE. Dar,1+lE. , staasocal nrdwde in cancar nesearch. Vol I The analysis ofcaeetcevol'sadies Lyon. Frurc: Internattonal Agenry tor RexarcA on Carar. 1960 IIARC uum6c publicanons no. 32.1. 21: Maclire M. Knt RB. Bryant MS. Skipper PL. Twreb.urn SR Elevred blood1kvcit of carctno:en &n pwive tmoter.. Am 1 iPublic Health 19e9: 79~1'3I1-4 . 22. Cumminp KM. M.rtello ~Sl. Malwee,r MC. Manh.ll JR Mr.bun•ner. of lifenme expasms oo p.une nnde. Am J EpdemroJ 19119: 130 I22-32 2.3. Couhar DB. Pwke GT. S.nst 1M Quesooanurt aaaourrni of lifeeme and nxent etpwuut to eavuorimeetd me.cco uoote Am 1 Ep+demwl 1919: I3U:'3311-47. 24. Cltm WC. Colpourec Ml. Fwg SC. Ho HC BnoscJyal cancer in Hbeg Koiy 1976-1979 Br 1 Caeret 1979. 39:1 t2-92. 25. Koo LC. Ho JH?C. Saw D. Aco" eahotinf .d p.wve enis" arront femde Iteig cancer paeno and controta Ln Hoeq Kong JUp C7m Gecer Res 1991,42:367-73. Massachusetts Medical Society Registry on Continuing Medical Education To obtain information on continuing medical education courses in the New England area, call between 9:00 a.m. and 12:00 noon, Monday through Friday, (617)1 893-4610 or in Massachusetts 14800-322-2303. ext. 1342. If writing, direct correspondence to: Program Registrar, Massachusetts Medical Society,, 1440 Main St., Waltham, MA 02154-1649. The booklev is free to MMS members, $5.00 for nonmembers.