Tobacco Documents Online

ASSESSMENT OF TME ASSOCIATION sETWEEN PASSIVE SMOKING AND LUNG CANCER 1/ARELAv LUIS R. DEGREE DATE: 1987 ~

This is an authorized facsimile, made from the microfilm master copy of the original dissertation or masters thesis published by UML The bibliographic information for this thesis is contained in UMI's Dissertation Abstracts database, the only central source for accessing almost every doctoral dissertation accepted in North America since 1861. IJ' - -I IDn'tormat on Service llrnversity Microhlms Internationai A Beu & Howen inbrmalan Company 300 N Zeeb FloaO. Ann AtOW. Moshyan 48106 800-52t•0600 OR 313/761-4700 Printed in 1990 by a.roQr,<phic proc.a. 3068 fiov%&C~~zo%

SJO r ~V CUM ordn Namber •6O1ee3 Assersment of the aisoclatfon between passfve smoking and lung cancer Vatela, Luii R, Ph.D. Yale Usi..nity, 19f'1 U-M-I }W N. Zecb RC Aan AMMr. Ati aq 106

sotZsCezoz INFORMATION TO USERS The most advanced technology has been used to photo- graph and reproduce this manuxript iirom the microfilm master. UMI films the original text di,-ectly from the copy submitted. Thus, some dissertation copies are in tYPewnter face, while others may be from a computer printer. h In the unlikely event that the author did not send UM complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copy: ighted material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are re- produced by sectioning the otiginal, beginning at the upper left-hand corner and continuing from left to right in equal section.t: with small overlaps. Each oversize page is available as one e.~cposure orn a standard 35 mm slide or as a 17' x 23' black and ahite photographic print for an additional charge. Photographs included in the original manuscript ba•+e been reproduced xerographically in this copy. 35 mm slides or - 6' x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact U11II directly to order. K 4 16 ! MT Atc.....0 n.. Nhnn~ MMan+.w.+.~c. fa"f~ XIONOrIh jWo /70W. AM Mpy LM ae/o6 130 VSA

PLEASE NOTE: In all Cas.s tho mateaal has bn.n 1Nm+A in the EM posyble way kom nr .v.i14,Me copy. ProbMms .ncounter.d with the documNq h.ve b.an WnWied h.re wth a check mWrk J. 1. Glossy p0wotcprapAs or papM _ 2. „Alor.d i4uitrat6ona,"p.r Or Print _ 1 Photopraons mth dark background 4. Illuipationa ar. ilOor cOOy S. Pap.g wdh black marxs, not OrnpnW topy 6. Print sho.s through as tMn s Nnt on (>oO siOM o/ page 7. Indistinct, broken or small Print on sa..ral pepu _~ 8. Print.r.cMds margin r.qub.nNnts p. : qn1/7 bound copy "^th print lost in soMN 10. ComputM printout papM with Indistinct p-1krt 11. Pape(.) LCkinq wfyn mat.rial rlG.iv.0, and not nraMab4lran khoot or author. 12. Page(s) ss.m to be missinp in numArrfnp orMy as ttrt follows. 13. Two pepa numb.r.d , T.st follows. 14. Curling and vwmkqd pap.a 15. Oes.rtation contains paprrs witlt print at a slant, film.d aa r.c.m.d 16. t3Mer UMI

8flMET,7-110% ..- % r,v.,,,. ASSES9MENi OF 'Dff AS.90QATIQJ 8E3WFE7: PAS52VE 94O!QPC AFO 1:94 O1tCR A Dissertation Presented to the Fbcvlty of tt., Craduate Sd+oo1 of Yale utivetsity in Csndidacy Cor the Mqree of Doctor of Philosophy [uis R. Varela Pay 1987

A2.'~?91T LSF53~E^. oP ThT .1S'0C'~1TIVI H=tN PASSNE RNt)IM .iND ;:1MG .^.J•=R i.uis R. Varela Yale University 1987 A :K;Cv..d asrcrntrol study was carried out to evaluate the associaunn Eetx+m OxpOSre to e+wttvrnc+tal tehaec5 saolce (Fess:w sadcusa) ary lLmy caneer risk. 'ihe study populatix was ompr!sed of 439 ases of lunq esnoer d-arnOsOd asQ+9 ncn-s^o~'1s. Ml of these cases .ero clinically ard histoloqi.illy oonfizeed. The oorzr_spo.ci~9 cor:trols were d_a+n fras the Nev Yark State Ueper*ment of hbtor Vehicles ard were :nii%nduslly .wtcnad to the cases on aqe, seu, county of residence ard previous 3Mcing history. A facr.-to-iace interviev ufty rl{+'_ied to obcain infocsrtim an es{+xun to emiramntal tobacco snnks. W inaease in risk was fourd associatel vith e)posias to three neesurments of spane sneklnqr or with eVosure to oo-.erkers' srohiriq.. wrverssly, e3posure to the sadu of othars in the hoaxhold was fc.ud ,.o affect th- risk of lu:q crr.caer. Fbr an exposure to 15C persanlyears ,)f ssnkirw the odds ratio was fourd to be 1.86. '11us effect aewas to be :arqer for epiosscvid ard snnll call tumars (GR-2.a]1 taan for a6--=arciriams ard other tuac" (OR-1.42). A+cza.suq mmoaa_-e to ;assive sm" in social sltustlAru was fourL' to W ir.sersely ass!clatsd with the risk of lunq nnoer. E~e iaFlicqtion of t.".is fL-duq -- at oAis with ~rer_xn results - is discisxci. 9-4 VOr' C,.C7LEf~O% AOQOW1EDC'~!~ I am qrent!y iJdebt.ed to Dr. pvSoht T. Janerich for providinq rte with the opfnrcunity to oarticipate in the pro7ec, frcve .hach the setervl for this dissertation was derived. rbrkuig vith him was not ornly intellectually atisularnq, but also a acairoe of qratifyiaq professisal arrd prrsona: interaRion. lb the M.K. Kelloq Floutiiation I am qrateful for ptovidinq a+e r+ith the filarrial assistance to cmduct my doctoral studies. Tb friends evid swbe=s of the Faculty of the Depaitmmt of 15idmdoloqy I am ;Jrrrcful for their .x.rti~unas help and supfort. tl

iv 7 naTA Co[2tzTIC[! ..................... 38 T11B[E OP CDrfly'lI5 8 FIISRIDf'Y FEVIIN 39 9 F2fPkT,a(,itC VARLAg1E5 40 Pal3e 9.1 SPOl6E b?U!'--DC FPBI15 ............... . 41 `Q1d.1l 1C I W-"El.'1S ii . .... . l 9.2 F70o6LRE 10 P1SSIVE ..................... 42 TABIE O P (XJNTTN15 s[7CaIG IN '!t1E HM6F}[7ID LIST OF TABIFS vi 9.3 PASSIVB SIqaQM Ild TFE 43 LIST OF ''I3MES x .4 WR1Q?i.1CE PASSIVE 5iap42rG IN 9OCIAL CTRCUMSU![:E5 43 C.`QPM pNE. 1 [.ITIRAAT L4tE A..NIF3W. 10 PV!'EMTALLY CtXlfQltdlIlG ..................... 44 1 BNCICIPQAm 1 VNtSdB1F5 2 DououxzC snrorES T 5 11 MALYSIS 45 ! 1 11.1 IaG:StIC RLG',tESS:DN 9M .... .. . 47 3 ORtffR STLL.IFS 16 .. ....*,,,... . Mi1TC7ffD GSE{i?lIIiDb 4 IABOPATJk1 20 SIZJDIFS - _ I!rf:STIC.ATI06 F7PtRA^QN15 *tttNl A 25 11.2 f30PC6ZOtE AS CJ1TVCT4tIUI . ..................... vARL18[ES m 5 . . 11.3 IXmOFiURE .i9 CtiMITfUpUS 50 6 SOIGNG AhD iCSSQ=C r'Pe oF [urc CxCE3e 26 VARI118LE ;0 QO\P".T7t 1~ac1E. 54 a1ArIL MEr.ioa ? '[hD. s. Rfs:JC.TS. 1 oFSQIIPr:oN or "4m sr.mY ........... 55 1 2 ^B.7F3 PIVES CAM'IE SIZ8 31 7' .......... SMMPS.E AND DISIRIB(IITON af TIO: lW1QiIDG VAR[A81F5 2 DIS[7tIB[R'_CN DF AOItMU1LLY 59 3 4 L=TICN OP Ttm Sl[DY FSI'ABL:.'ACgNP GF T+ic ................ 34 34 .................. CQNFQMM VNRUBLF.S ) AS56.'3QNf or TfQ: EFTTZTS .................... 1 C.\ST5 aEPCRi'I?G SY'"IfTI C7F PASSIVE 9'UFSlG ON ILTG - S .1 ELIBII.ITY fP.I'[FiiI11 ..................... 35 C.lllE3t RISKS FJR CASES 6 !M! CF 07NIaDfS iii

9L ZL L9 99 BS auod TA 'S2KM.t 91 .10 '.]1FYi6 35tDdS QL '~~m JO 7YLIK3Kiia70 V Uw &:.I1II'I 3.it?OZ.$H[D QtV •9JISYtl SOm S3LY4Lwm Qus1!loVNfl ........................................ .9 Q1 alC'RL ' 1 S 13SflDdS 3}il A8 2mIOS SwQI• a:) 1fl8" jr) SdIonL10 )QS OL 31i190rflEi N[)l S.1.I3Q'I Z-?Lalal'n 6S6 Otv 9DLLVtl 9D]D •ST'f-tLLq 'SClIa19RL 'JEWlS3T*3VVD3D OZ .iO `.7MOS 3S[1p3S QL 3U1SOiMG aLl 7YI.IFIIKiMO V liQJ Si.Dg'I 3~.~iT7 Q1V •SbLLVtl 9C» •S3116aiS'i O3LSTlC'0V .......:................................ •Y Gd ai9RL 'ava/S31r3a10n ot ao UNnns 3sn3as (sI 3:nsodX3 ao 'rf* 7431GAZ a v an~ si.ua-i ~~ ca~v ' eorrn saoo • san~ awsumem .........................::............. .f pi ai4w '35<TJdS 3HL 1EB aMS xW!lfid exrLTMD 30 53rd333M 3AI, QL 3915LTflCi 1ii SLDII1 3DN3aIJNM tS6 OrNf SDLLVtl 90Qll S3L"'dLI.9 ........................................ 'Z a11 alQSL 'tOLLYII131d Aflllb 9j:L FII S"9YIWA ilMTLStl A) 1OII[)HIXISIO ........................................ •t ON Ol4RL S31810 JO 3SPI S91 9Sl ..................... B Z l ..................... • MC1l Ki.1dt'}o dll Ze EL Z9 Z9 : ;ji+A~4:j 2023382411 'nrtxs ~ Ni ORnias 3AISS~/4! oL 3L^ISOD3 -'t =Dt'WH]M 3K:. Ni OPMIO6 01 3K'L9COC C' C •'•'•""'..•••"• F NDLi&IlSS•t K~0~1.'1 SPIIIM Z'Z't ..."."."'...:.. I NOI.UNtSSY t©Oru1 S.L'IILS3a l'Z'C Q701iiS:lOH 3Eii. Ni '"'L')Cib 3AISSYd Z't ;3:I C3IttdV:: DN:ZiW 3571clS G. Afi M17~S • .•.•.• . .• :•: ::: • S3LL7h,%!JI3 dD 1Q6iY1 :Yd= C' l' C ONM1C1S 35TCxlS 37 StlY3i. d0 K.'Fbl]N L' l' C 35d10i.S 3RI AS SM6S .lW>S'sL:.?dOI) 30 1a%i1N 3Ki. d0 ,;ta-S53SSN l" i' C ?SfIOaS 3K:, d: S.:.iB`1N !ITYCoS i' C

-StrDIS1Q1fSS 7VIJOB NI 3KD10w6 3AISSId A) SM-dTnn^ dffll3 0d. 31CL90'DCd llW SSLDd'] 10 t OZt 3D?aaLDw K6 aNV 80LLVtl y0op '~VFQaSi ..............:...::..::..:::.::::...... .oL eN STqRi. 3DIfIdYlt?7 3il. NI MLDp6 srakNosx3d ost OL 3)fL9[W .Ib 'fVLiXraiiIa V aol S.ISHPi 21ana![7D CNU soIL1H1 9t2b 'SaLMdi.L'.R a3isr1[Uv 6 L[T ................c...c................... .6t aH atqR6 30rlM/!QI 3>}I. NI ONIM0S S1Pa&/NOSlpd 0st c]L 7a19c'.na JD 7VI31Fnina V aai S3.DQ'I 3343MJNpJ aNV SDIiYtl `.i00p 'SdL*IIj.Si a3ISf1fV1RY1 6 tIT 'tI wJ aT9RL '33IrI&DXM MII. !u 'JNI]IO~s ADsincoaZYJ_.13DI3 aJ. 3flr19lk0o Nf]i S.Lnd7 xwaaL&au ts6 aMt soi;Lvu saao •=YruLm 6 ITT 'LI ON aIQRL 'a'IlFGSii>f1 MLL NI n1tn168 511KiJ1/NO6'Kid OSt Di. 31l190rDCll A) 'IYL1KDliilQa V HnJ S.LDQ7 3=1.33Libm QIV S7LL1f!! 9QYi 'S3i16tILSi QiL41[VY B 601 'sT cai at9RL 'QIt7fl5fYI1 3i1. NI 9t>D1CSS A) Sxrwo=YJ SiD13 UL mff19C1flC1 NU! SS.I)d'I 3=fFldi.fib tS: aNV SDLI.V11 SaOD 'SII.VIQ.LST m C0I ........................................ .tt aN at4*C1 I tB ........................................ '9T ON a14RL ' a'gfl5l)!31 3R.L NI :HDI0W6 Sb'ZLL4x)6'Im OSt QL atlL9oem a0 7VLUlDMWa 1/ N[i S.LI14M 30KMLYD aNV 9DISYtl 9qfD 'S3L1bB.1.,3 DnWdV1R7p 2023382412 '7IaF3sfnH 3IU. Nt JNLICYs SWLs./4nsird o€t ot. 3ansaoa ao 'rr..tN3tm+na v aw SLI1tI: 3:?~ Q1V SOL3.VG Sa00 'S'.aLvtILSi a3Sf11t7V ........................................ fI ON aIaRL C:oH3rfX71? 3KL NI '..`VDIWS SZl1QJ1/N06'1[fld OST C11. 31Ct9e><CQ j0 WUNM'1itT0 V t!(.i s3.IbII7 3DN3M&M OtN 9L`Ii1fY SWO 'SiLw6= a3.LSf1tU1QU` ........:............................... ZT 4d olQt[. -a7oeSrr.H mtL NI `JHnIOS .i0 SISSMQ-iYJ .1H)I3 C1L 311119;knG NCIJ 53:.'Sd'1 =N3mjNm tS6 Ow 9ULY1f 970L •53i.VhQ.L. ............................::.......... .lt ON otaRl. 'S3LI-3dRJTr 000'00i JD ~'D#CkS 3stnds UL 9•en9rkt'Ci do 7KLtIZi1Gum V aal s.taZII7 TSaM.laloo aNr '90Li.w s= 'S3LVla.ta3 a31Sfltvv ........................................ .ot ON ataaL 'S~dI-"r+VDD OoC'OOZ AO 'a1D16+5 35fk~3S OL 31[15T~Ci m 'R'u*,-sdiiLfla V t!w S1DM 3:?LiOLIQiD aNY 'SDISYtl Sa6J '~'iai*aJS3 a3i%l[T]1Rm ........................................ •6 ON atQRL 'ZO'3 2iMM ?1Itlf10 IS13SlCl.S mi: 1l® awB S3S.=HF.*'D dJ iDFSrw 7Y]aL .r. SiIHM;tYJ 3AId oi. 32nSOMG a0d s3.IivL-I s'23ma+rD ts6 aNC 9oL•.Irc SOm 's3LvPa.tS3 ...............:........................ .e ON DtmeL 'SbY'.,.t Gt JD `.S47d1S 3S'10dS OL 3kl19?flCd 30 T:-Jh'mt+ta V tl[la S`21II'' 3x.'IacatCJ ftflf 'S LLVd SIQJO 'SSi.16Q.l.aa GLi&MVY LQis1CRL tTll. Rn

LL iL OL 69 m ab.d x aS^SL1'~C 37 S~P1G~'~I.~ ,,rn.,:~~0 e:d s0].J =1 cNY ~sJlz[NOds31i y..ffoilifls o+r •S.'•rL1Q"s3d=.a-L' •S.LZi.-er1S TTV ani 'aMos 3SndS 3u. SaVaa Jo xmrY+ Kur+ smo an x=NND Urn ru Xrirta -ScN am%67j '31l79DiflCi d0 S3IUOFfl.LrJ dDQifii'10 " 90LJ 7C7 OT.V =33du DIDCrll1LSIA iL~i1D/VFI71LtlVrJQ~al ~W! TM T1NS/(UdWflC.Idi •SL'XilE(1S TN tl0U '3SYliS 3iL x8 CMIOS 1CVO 8d ga LCi~D jo IE3wIYd KLIM SOm a77 lCLNYJ ?UT] NI MMD 't aMarb1.i ' 31ILsam jD S3IionII.IXJ J"Mia?U 1Ai SmD 9Ui Qdlf %Sr]Yw WY S31YKi,i - 'SL`x3f'6(lSi TIY " '3SfK1lS 3Fil ,TB ®iDC JIW K3d - SiLBWJD X t[iSiH 1tidM 9om JCfl Kitr1D JWTI HI 3nMD . ( c" wm%bTj '3Vl9D)Q M S3TdOfZXM*lM3JM0 UQQ3 - Sm7 `JOI QtV t SULICW K3t" QiV SKi16S fflAa1 'SI:YifBtLS ?T.' tS..^.-i •3MfJd5 MiL .L8 ®td6 AtlO t©d S3113dwoD JD i[iif[1H My IM 9tT10 :70I lCi=li1D SNM fiI 2SNdD ................................ .tCH a2nbtj '3K)SLYJQ rD SarHOUZ*YD .UiMMa iidd SQJJ D07 Q7V ~ S,iMa4DdS3!! 3S1f-lOtRpLS ON1/ 'S.L"UiS3U-d`LiS •S=rWIS TN 1!(li '3C1DdS 3F31• L8 Q3AOs A1A Kid SMLLI1f1OL7 AD t80filN BLIM St70D aOI ffiaNYJ !INR] FQ 3DWdD ' i ON 7n6TJ 31MIA S3Y17Id.AD ZSPI 9ti iti 2023382413 •'sII'yCCHS 7AISSYd 7KLYJ6 O+L 3ltns0m d) xnII 3RL HI S3.IW1 Ot d4 31RMM dD 7YIJKAIIiM YMI SLDM mKmL*tD Qw SD7I•Ytl 917m 'SzLram OuSlf9m ........................................ •rr cM: oiqn 'brmos antSgVe •rrroas nt 3WISOM M z33a 2Rt ru St1z+o sr AD 3aunM Jo •trJ~a V MM Sr.uun MNMLNM oNV sou.vu 9oao 'su:u.13m XU-4uv1?m ........................................ .Tt aN alv*3 XT m

n I 8 YW 6 a 0 N Y. ~m f~ F3 ~~ ~ ~ r Q b 0 ~~~ ~ ~~~ 13 Elio

N M' N E {i 0 :l g 6 F. N N 2 is'~.U n R i5 . ~ a3 r 0 9 ,.e i Y0§ IV ~ ~(~ ~ M Jz Z L11/ l/Y. = G u o u i 4~~ o ~ C1PG; T A

aH1Y1lR @>E LI:IIU1pRE 0...VIB4 1 BJ~II~. !n 1972, tthe AepaR to the lktitad States Ca+qr'eas on the iiaalth Cunreqxnaes or sndcinq fgc the first time gave serious eonsideration to the Sssu" of the effects of envizrlwsital ta6eoco scvke on the rron- smker (1). That otlrrs besides ssokers ttrieselves oaild experioWor det.-irrc^ta1 hrlth effects due to tobecro ssoks Aed bean in the sunis of srwy for at least a deade. Ob.tttticians in P.rticvlar seeald to haVw ilitQlsifl(d this prlQ'CSpetioll, by rlporti.'ttj tMt cigarette sedcirq aeQng preqnant arnen .as associated with pramaturity (2) and with 1.a+ bi~th-reight (3). Other repxts attributed a wMle array oi adverse effects of s.ternsl swktnq on the foetus. ranging frae i.iosired pos:-nats! develornnnt (I.S) to cxroeai,-3l snlformaticns (6). Iwter studies atter.ded to the eff-eTs of tobsczo oomponents released unto the environnent at /arqe. For irstsccc, studies in the Cniced States (7,l1. Fhqlard (9), Lsrsel (10) and otlrr cas+tsies /111 sf7`ef: 'IOM '_Ri1flan of sA[iCif1Q Qerelts had h10f1er i.'K;,iPJ1^_t :f .l. -2- res, uato:/ infections. as we1i as tur,h.cr hasoital admission rates. [aboratory imestivstiaru showed tlAt espmca-e to maternsl snokinq prvduced a defective devrlopnent of puLmwry fvnction in children. as ..,asured by foroed expiratory wlume in ane secod (12). Finally, reports of childhood tusrs assonated with p.rental ssdcinq alpeaced in the sediol literattiue, although otten with oontradic..ory results (11-17/. :Tut there rere senom he.lth effects an the adult passive soker - such effects as thre experierwed by the annkers tleemselves, i.e. lunq uncer -.+as rot considered poasihle as reoently as a decade aqo. Triis an be gathered fzta the follovinq paragraphs published in 1975 : 'In suanacy, the effects of ciqarette ssdce an healthy rnrt-snok.rs oonsist seinly of sdror eye and throat irritation. Ho.meer, people with certain troart and ltmq diseases (argina pectaris. ECPD, allr:gic astfa.) wy slfer emcer'bstion of their synptose as a result Of eqxoaue to toraew usktfilled enviza.enta...* (16). ' ' with respect to lvnq ca^c•z there is no evidencs to indicate .hether or rot this level of expunae has an effect on the risk of developing lung canaer. Ho`+ever. tecause of the low dosaqe and brief espawre, it tiould sean unlikely that there would be a siqnificant inQeue in the risk of dr.elopinq lunq canc:r " il9).

-3- It is an establishd fact that the overall cancer rist for r+m- mdaers is lowe.r than tAe cancer risk of the qeneral populetion (20- ZZI. Crnnsred to snokers, norsmakers' risk far mnoer is ahcx,t 50s 1o,.r1 (.C1. Ft: orrtain sices -in partio.ilar the respiratory tract- the differm~ces are even mare drmatic. A recent study, noxtheless, has suqg--vred that Iunq cancer asQy oorr-~ way be on the rise. 1LTe suqqestion of such an ircrea.+e asonq non-smdcers has tocti+ed iaterest in passive sndcinq as one of its possible ouses (23). T1Ls suq;estion is partly based on a better uder3tardin9 of the ca.^cilnqenic qualities of hoth aa:nstrewe and sidestrami cigarette snoke, as will be discussed at a later point in this chapter. thstrnn (23), using data froe two ceprese+tative sasples of lunq cancer rlearhs and eortality statistio at the national 1eVe1, foknd tor the peziod between 191{ and 1968 a 1S-fold inczasse in lus+q cancer dsat.h rates asory rhite norLrokinq srles ag.d 35-84 y+Nnrs. The increase .ns the highest for those srles over 65, estineted as 30- fold. Fbr white females the increase has been 7-fold for tne 35-94 aqe qrt.;p. 'Tuse changes seee so drss.tic that the possibi:ity of error!- in f}utzan's est'-r.etes has been r.aised: same of ~Lis data 5ourcts go as far back es 1914, ard the_e are no means by wmch to cherK t'riz :e;iability. 7here are also reasory to doubt tte amuacy of the ascer=airzaarK of s.rokin.l status in the surveys from which tke data was obtained. Moreover, tte categorizatian cf the snDwA •r}x:aGie cl'JllqRs tr0111 o11e sa1Zte of data to aflotfY:, as does t-he promr.:an of .rlf-reported snolcirrq versus rir-rqate reportuiq, tnus fratLn-i the se of pocled ?a:.a difficLil Lc u1ter-,ret. It nas also cr~l+ -•- pro(nsed that sarne of the changes desczibed by fhstrt.n may be actributad to better diaqrnees andlor dunqes in- tiaor classificaua+. Ibyever. as the aurJor aaa posnted out,, the tzenr,t persists after 1915, rwhen the sa)or changes in diagnostic eriteria hsd been iaplenented. fYnally, ttrre is the po.sibility - as Fhstrts suqqeets - that the rising rstes indicate the irur%ventian of otAer fact..as in the causation of luiq cancer in adlitirn to personal cigaretrr srcycuq. Other researtfiers do not agree vith fJVtsam'a findinqs. Garfinkel 1211 oorpsted lunq cancer mortality rates for rnrt-smokinq participan•-s in two large prospective studies. In the Aser-imn Canwr Socety Study, 94,000 nmt-ssduuq wales ard 775.000 rnrrsindcinq feeeles were enrolled betaeen Oct.oeer 1959 and r#sch 1960. The data used by Carfinkel spw the period between the erd of ttr enrollaent ptase and .7une 30. 1972. 'fhe secvnd study, the porn Study of Vetesarv, wes initiat.d in 195s with the e.ilinq of questionnsires to 293.000 veterans. Fifty foLr thousanis rezpon3elrts repore.d thonselves to be nrn-sAOkers. 2he follow-up period anded in 1%9. lbrulity rates for ron-ancken in both of the qrvu{a studied yere est_mpted, msinly thtvuqh the use of death aertlfiotd. llates were adlustsd to Oe United States adult population of 1965. Garfinkel .as u.-sable to detect any {••^re••• in .nrtulity dur to 1u+g cancer tor 1'.ree four-year periods between the years 1960-1972. MeverLhe?ess. he re4,crted a sliqtt difference ( alttaouyh not statistiolly siqnifieantl in lunq cancer risk for rcn-..dcinq womaq s.rried to seds*rs, as caT-dred to nrn-ssolcinq .+aen srried to nor-ssdurs. Zhis .as probeoly the first pooulation study to oonsider such an association. ;Te results will be discussed in t'rse rrwt sec+`r of this ctrpt;-.

-5- 2 FpIDEMI0IMIC SiVCIES. C.rnzvuersUl raports on lunq canoer trend+ auusl rm-ssnkers have :rot beaf disomrragin9 ernu9h to dismiss the nypothesia that passive sudcing -- also termed involuntary saddnq or secudar/ smc9cinq - it causally related to lunq cancr.r. Ln January of 1981. the Br`tish Medinl Journsl pti:bl:alyd the first pnpulatian study spenfically addraasi'x9 this issue (25). 8iray.na reportsd --'w fiadings of a i1 year old prospective study t.+udh imvolved a follor-up fas 91.540 nort-mdin9 s+srried Japanese wassan. Ckw huati'red and se.erty fau cases of lunq cwxer had develoosd in this qraup, for which the hiuhard's s.oking history had bee+ c.vl3ected irrieperdently• women mrr:ed to hsvy s•vkars />-.0 cigarettes/dayl snewed a hiqhrs risk for lurq ca+cer than ams+ sarried to non-ssoken (standardized risk 7tio ~ 2.OA), FUrrhersnxe. a statistically significant dosrzesparue relationship was foiad: the relative risk for .rives of eu-sckers or smdrers of 1-19 ciqarettes/day .as 1.6; the ze:ative risk for heavy 2 ssokers (>.20) ..s 2.06 (Iwntel-extaruion X teyt - 3.299s p, tw tails - .00097. The trerd was also aoserved rAw+ huabsrds' age and occmipation was takan into aeoont. the hiqhest relative risk. 4.6. ,.u found amaq wien in aqricssltural families aarried to heavy srokers aqed sC-59 years at :im¢ of fttrollment into the study. N&> .e+sse in risk for other na;4c casoers %.as o~ser+ed i:n relation t?e husoanas • snohinq habits. Ri--syesa's atud/ rectiv.d a great deal of atcr3+tian amonS ooth t.he necical conti•ruty and the .ay pibi;c. His results ari: r,et.'xads -6- ware cloxly scrutinized. pramtuy an eichanqe of letters either aitical or sspportive of the study. •ltre fact that the Mantel- --__ - extension test tor ors of the tables was calculated errvcLLsly was particularly eaqhasised. A svbsequent estisate ob*a+ned by rearranging the data was publiahed suhsequeritly (26). The nev figure d:d not zharye the basic eoncl•sion thst the study factnr and the disease rere significantly asso-+ated. This resssu.-anae, rwnetheless. did rot preclude eiticim of the study on ditferent qrtRSids. first, the research .ork - started in 1966 - did rot initially intend to look at the pasaive snokirq-lunq canoer associaLlm. This parzicular :mture of the study ssiqM have affened the quality of the iSora+stion in several oppoaite .+eys. an the one hand. it sey be that the likelihood of bias in 'reporc.wq' or *irrtervievirq' was less th.n if awell-denelopsd hypot'rsis was being testad. This is because the study sub)ects, as .ell as the intarvie.ers, yere obviously blind to a hypotAesis which was non-eci.stent at the tine of data collection. On the other hand. we should caLsider that the research. rot initially intended to study the effects of passive smckirtiq, say rot have bem as thoraqh in the asaer2ainsent of eaposure to passive saokinq u.ould be necTssary to detect a noderate or snsll effect. This latter point is not wrrisare. hoyeweer, since it wu:d have resulted in a coaser%ative estimste of the effect. Secondly. in the oriqiral publiotirn by Hirayams no details are qi•.en on the ,..y in %wticii deaths rere asoertained, nor ras there infocrostion availaCie or. •.tie degree of pst.holoqial eonfirnation of tne .:ancer cases. lt is only rtuwr,i; subsequent puu' lioticns that we to

- r- leuT that in a sul~e+nele of 23 cases, 17 .ere adenocarcin°'°s• Since smo1 yiy has been postulated to be associited with specitic histolagic types this informat_'.x: would seem eucial. Oespite >.he drav0ecks mc:doned atvve, there are oeruin :estures apm.-t Hizayams's study that swca it uniaua and prtb°bly very difficult to reolicau_. The prasPective desiqn Imiolving a large poFxulation in orte ti:an on geographic reqion is certainly an astet. Also adyantayeous is the faR that the atudy wa+ caiTied out in a traditimal sodety wher'e ssoRing habits of wnen are 1oV, ttus disunishiry t.he probsbilities of ineludinq erzu:eously classified women who developed cancer of the 1:s:q as a oonsequenee of their o-n smidn7. A1». i.., suGh a setting nsrriaqes -y last ionqer, houses are certainly mmh ssrllez. and socialitinq for .+men usually talw pl:c! ir. the cznp.ny of their F+uslaard+. All of the above suqgestv that quanti!ication of the hushands' .ukin9 habits esy inSeed be a 9o<+d aeasure of e:posure to seccndhand ssoka.. Also in 1981, tYichopoulm et. al. publiahed the prelimin+uy results of a study eonduct..d in Greeoe 127). Using a ase ccntrol desiqn, t.-ey asseabitl a qzdip of 40 nars:cking waen dia9^osad with i:arq cuxcr in three large Athens hospitals. The patients were inter.•ie.ed reaazdinq their huehuds* srokcin4 habits and were ca~red .:th the husa.nds ` smokinq habits of :49 nort-adun9 hospital controls. The centrols were octtxpedic patients rho did nct differ significantly fro:s the oses with regard to age. ediaation. c1:ra`•;on of mnr:i3m. occvpstion or resiftncy. '_fie odds ratio for ween c:ar.ied -e- to ex-ardcers ..s 1.8: it w 2.4 if they ..ere mnrr-ied to currNr.t sflokars of 1-20 ciqarettes/day; and 3.4 if srrried to current srtokers of snre than 20 ciqarettes/day. ihis dose-response trend w fourd to be stati.stically significant (Cu-sq for linear trerd - 6.45, p(2 tai 1 s) < .02.). rtien expostue to husbsnds ' sanking over the oourse of the cnrriaye .as assessed a similar trend was observed. Pstinated odds ratios veze 2.5 for tlose exposed to 100 - 299 tlv;sand cigarettes and 3.0 for those exposed to 300 thou+ud or cigarettes scre ( Q:i-aq for linear tzmd - 6.50. p (2 tail) ( .02). T.us linear trend .as idential in both qrow of cases: those with citoloqical emfirmstion of tuouar and throse with clinical diaqrnsis only. A later r.p- t of 7YicttpaAos' scudy in 1903 added nearly as m.ny cases,, and 504 sars cu+tmis, to the study sub jx.ts (28). '!he results rere substantially the sasa: odds ratio of 2.4 for wives of smokers of less than 20 cigaretres/day, and odds ratio of 3.4 for rrnen aarried to snokers of sQe than 20 cigarettes per day ((hi-sq for linear trend . 6.7, p(2 tail) - .01). Tbe Greek study, althouqh not entirely oosparable to that of liirayasa, shovs sisalar results. It alw shares some of the adyantaqes of the latter, nasaly, thase mrcerninq family structure and family sociel life, together with their isvliotions on the assesssent of expnu¢e, as ewsmerated abav.. Oiffarer.t srthodoloqical problens, lorever, ought to ne :znsidez.d. tbr instance, the study qrvw is siall and the dcqrx of patholoqrol oonfizsetian poor (only 654 had a c,toioqical or dtis.aloqiol di.rg:osis). 'ET+e study suojects (cnrr

-9- sno,uaq •casen with lung cancer) reoresrnt anout 781 of all the female iunq carcer cases diagnoaed in the participetinq hospitall. As is menti,=+ed in a subsequert. see.ion of uus chapcer isee: liistoioqic :yae sai laaw £ar><u) the propoc=ion of ncxr-smdc.uvg patients in oclrr series of luy caneer cases has been +stal,listsd at about 10% for %amen. I!he exoessiveiy high proportion reported by '1=rc)Ppou'-aa suqqests that some of his r•ases way i.-ideed have teen ssukess Who did not aini: to their tobaoco habits for social or otiY.r reaeoro. Or. altrough rne propoaitlon is saeswat unlikely, it wsy be that r+on- snok.:_ba wden with lung cancer in Greece are aore likely to be hosoitaliaed than in other parts of the rorld. In ad3i:lon, the fact that both case.s and ooxKrols were inte-rietied by a sinqle, unblincled resear+cdr.r, points todard the possibility of uisclassification of esposura. Ctn the other hand, the attenet to assess ex?osure actvrding to spouses' sodclnq haoits during the tot,l length of married life sry have provideJd more ca-+plete ascertainment of expouua. Pelayo Comm at. ai., using s case-m+trol design, assessed the risk of lung canoer for both non-sadtinq sen and WWrn .+ith reqard to the sadci nq habits of their soa»es ard parents (29). 7Twy (ound a.i adds ratio of 2.0 for 9 ssm owazied to wt-.w %fio staoked more than one pscl: of cigarettes a year. The correspom'in9 odds ratio for 22 i.crten Tti:_ed to rdcinq sen as 2.07. NaitAsr was statistically siqnifscant. After oo+trollinq for the spouses" sndcinS h0-'51t, msce:nal (but rot paternal) smkinq %as assoustci w,.rr a hia:tr r.tk ozMeezoz -10- of lury cancer, as indioted by an odds ratio of :`1.47 (fpo:.site resul:s on this parental saicinq-lunq nrxar association have been reported oy Sandler, at. at.. 'ltmee reeults will be discvssed latter in this ctiapter). 1fr effect ebearvad, lnre„er, was neither Isiqnificant nor seened to be doea-relatad. ftwm e~,a>sure to both parents' sadcing was embined, the adds ratio becase siqnificant only for t_tiose mses •xposed to - > 41 padcs of cigarettes per year (CR . 3.11, a < .05). Iaportant drasbacks can be !ourd in the st-dy by Pelayo Cnr±ea et. al.. Its sain disadvantage is the faet that the aaber of participants is so ssr11 that any of the poasible strenqt.hs of the study - such as the 100% histologic catfinwtion of diaqnoai.^, - fail to carpensate for this weakness. A more recent epidenioloqic study on the lung cancer-passive wolcinq relationship was carried out by Car'fir9cel, Auerbadh and Joubert (30). 'lfieir qroup of cases was csaprisad of all !enale noi- s.okum lung cancer patients diaqnoseE in three hospitsls in Nev Jersey and one hospital in Ohio. 1raim diaqnoswd with colon cancer se:,ed as controls. :T,ree controls -ere satet+ed to each case on the basis of age (•-5 years) and bospital of diagrosis. Patholooy slides for both cases and =ntrols were reviewed blindly in order to ccnf:.rm the diaqnoeis. C'ne hiandred and thirty four lung canctr cases and 402 mntrols were then Lmemie.ed regarding the snokinq habits of their hLLsbands. Current s.i a:inq habits, nLmCer of cigarettes ssdeed at )--, and n:sroer or vears a-e hnisbands had ssoked %ere assessed.

-ll- Su:ular!y. wmn %ftre quest:oned about tne nuwer of hours a day they Aid oeen expn4ed to the snoce of ousers in the past 5 and 25 years. It wns found that womn siarried to ssdcers of :0 or nure c-garettes per day hut an odds ratio of 1.99 (95% C3.:1.13-3.501 as cwpnrod to thore married to norrsindcars. If half that asourt of c:qarettas wea smnked at hone, the adis ratio :ns estisated to be 2.1! (95tG:1.13-).951. Simtlarly, an inrreassed risk was found for t'ose „o,en married to men who hsd senRed for 20-29 yoara (OR - 2.2), but. ione:mectedly, rot for thoae swrried to ssoloars of .ore than 30 years. No inCreaae of risk •rs found to be deperdent on the rsarbnr of e:noeure-baus in the last five and 25 years. l:sinq logistic r+qression tsctrniques, a sodel was fitted to t-he sr.dy data. The sodel included teztn for po.sible cv+foia,ders and sev-sal eontinuoua variables representing exposure. An increase in risk vith inaeasinq espoeuae ras found for the nvMxr of cigarettes snoked by the lxisbmd at has:. This trerd ras statistucelly significant at the 9.05 leuel_.. The study by 6asfinkel et. a1. attesnpted to tadkle sone of the aethodological prvblene present in t.he forse: stud.ies on psssive spo:,inq. '.tie authors rere suctessfol in putting together a sizable y_o:p ef cases •tno were cvi-`isnd never-a+aicers end whose diaqnosis was h:stolcqinlly ccr.firred. 'lheir aabessnent of expOSUre %._as ner. lunitad to the snace deriv.d from the hustands' mcid.q but included ttiat er,rrlex.r.d at tltis wrksite and o!..hes 3rees. !ts- r'Jl statisticnl tec'~niques (leyistic reqression/ were used to ana!••ze c'w dat-a. oradLcuiq results c=ss stst wia the f:n3irws cf c-ie rw : Yc -l2- ms)or s:udies on this topac. It sems, thecefore, that the only asper of the study that eould have affected the results is tnat wh.ch is valt difficult to evaluate: the reliabiLi'yo of the data collectrsf thrvugh the interview. FWm tfusgh t}w i»ervie.ars were spssedly blind in raqard to tfr ifypothesis of the study and the case/+xntiol sutus of tne part.icip.nts, the partidp.nts tt~elve, .a-v awre of tlr.ir o.n crnditsan. In this situation the pmailxiiicy of overrepor*-ing by the cates sust be bonu in eur.'. Additionally intervie'Os of surroqate respondents sy have affetted the reliabi:ity of the information al:aut the expomae y.rjable. In faR, rantel (31) has slo+n that Gsrfinkel 's estii.tes of thm adds rat.ias vary oauiderably .ccordinq to the type of zespmdartt. Be found t_hat the «Hs zatio • for lu+q onczr w].S7 (95e a.: 0.&1-15.29) in the subqroip in .4tiich the respandents .+ere eitl1r a dax7thar or a am of the study subjecu. In the s,bgitup in .hitA the husband pitrridsA infornaion on the exposnse he found that the od3s ratio w 0.77 (95K3.: 0.56-1..06. Ard finaliy, in the suEq+roup of cases wio report.d about their o.m ex.nsure, the odds ratio vas estiirted to be 0.e3 c95t Q.: 0.56-.1.3e1. At the ssne t:me that this report was being prepared, tw addititnal epideniutogic studies claisnnq the exisctence of a soderste effect of passive arokinq on luny onoer risks .ere published. The first of these .as also aordkrtad in Japan 112). It is a o.se- crntrol study, but, as vas true of its predacesar, it cnkes use of a cohar: •_litaally defined for arut_'Yr purpase (the study of the effects of a,cmic rsdiation in !ii_oahl.-s and Wr)l4su surY.vors) in orC1r :a

-1.- -1~- aettnd,logic problem that cast doubts over-their results. One ~h obca.in its study sub3ec_s. The cchart, ornpr=sed Of anout 110.000 prmlem is the poor rate of histologic aonficnetion 1Te authors in . , ;xnplz, yielded 525 a3ea of lung ':anarr between 1971 uid 1980. trying tn cirtumvent this prnblem, carried out separnte analysis for Patholaqit diagrznss .es'e dcx in 579 of these. Controls we-re those with pethologic diagnoais, as well as those r_rhait it. An sele ted from tne eaoe cnhort aaang meRxrs who had rot de+xlop lumg uodcired cviS equenoe of such appro.dh, harever, is the loss of canorl. Tt+e controls were individually mstdrd to the cases with statistiml pwer due to the resulting serller gru,ps available for regarc to year of birth (+- 2}2a)2 residenoe, sex, vital sutus, and Ertiew-r or ncx t.hay had belonyed to the cnnoc: subgroup tlut unien.ent Sierni4l medical esamiltstions. Inforn+tion wes obtauned for 428 cases and 957 controls. Horever, only 13 e.le easec and 94 fas+le cases ocwrred anong r.on-sroks_-s. Aso+g the ron-ssokinq neles the odds ratio associated with hsving a spuuse vho smoked was 1.8 1901 Q,: 0.5 - 5.61. Fbr feaales the mrresPondin9 figure w 1.5 (90% c3.: 1.0 - ..5). in adiition, feaeles showed an increase in risk of 1ia+g caneer with the inQeasing rxmber of cigarettes snuked by the hus6eud; the odds ratio oeing greater than 2.0 for the hignest espowrc category. The test of linear trerd, how+er, vs not statistically significant at the 0.05 level. Likewise, arother seuure of husbands ' sndcinq -- t•e nurber of years the huabsrd .~d~.d - was rat stati+ti=11y significant. rlnslly, their results sha.d that nrn-ssoking uooen who had beer~ e~osad to the hwEsrds ' ssnkin9 in the last 10 y.ars had a lo.er risk (Gt  1.2. 901 Q.: 0.9 - 2.s) than those ram+n exposed within the lsst 10 years IOR - 1.8, 9ft Q.: 1.0 -].21. Sae of the .strnre estiiostes are consistent in their msgnitude if not in _heir s'•.atistimi significance) with the results of ene l:rst Japanese study, as vel: as with the ot;ar scudies on pessive acweve.. >his subxGuent sr,ay aqually shares in sane cf tt.e estimatiom of the p.rueters. Arother potentially iaportant prablea, cmoerns the obtaininf of the infa®tiot about exQoeure. On1y 22 ases and 26 rsntrols out of the total, alrhouyh not necessarily ovt of the naR-sioking 9roup only. were ava:lable to be intecvie.ed. D=sure for all others w assessed tJacvqh the intezvieving of surrogate resFa~derKs. E14- tJn.agh the distributirn of types of ssar>qate resportlvrts .as suiilar for waes and eantrels, it carvsx be ensured that selective saeall was ellatinsted. 1. final point otnenrns the pecularit•ies of the study gro,p, navivors of the atarndc erplosions of R+.*+stiim and Nagasaki. Sinw both cases and cvntzols were equal in that regard. we do not question the internal validity of the findi+t3s beyvd the srthocblogic problme sentioned. Howeaer, - should be cautious in any atteapt to e:trapolate these results to the gevsal ponilation. It could be thst the effect oE.ervecf is only a praduct of a synerqistic effect bet.een radiation and e+posure to todsccn sndce. "Sis syneryy, of course. .euld not be posaible in the 9eneral oapulatien sim on the .ho1e it ladcs exposure to the levels of radiation esperiencsd by the subjects in this study. Ilnaily, the authors of this stvi~j have been less thsn aonservetiJe in esta,blishittq that level of statistical significance rhidi they are vill" to act'ept. Values of p betw+en 0.05 and 0.1 are reported as

-15- signifioar4_. NAe1ti.xally. 906 crxL`idence 1'.pits are teoarted instead of the mr.w+tloK1 956 CL.• Althaqh fsaa a nethodoloqic or statistical point of view these is aoth:n9 v=y with this approach. it raises af+oe-'n that resea_-chers, i~ ortSer to detect and repnrt 'signifiant' effec_s. sey take the risk of aaLvpeu+9 a 4re°ter role of chance in their !ii+diJqs than that Quch is convent.ionally deemed apnpprLate. The se= d most recently published article is Sa.sed on the pooled data fcae three case-ccintral studies (3])._.Ninety nine histologically coniinR+d lung cancer cases and 736 controls cmiprised the study qruw. these .ere all the pazticip.nts in studies carri.d out in :y;;yian.. i~.~as and Ne.r Jersey vto rt=ted thsaselves.to be never snolaers. Oue to the fact that the data case from three different sairte-s, the type ard level of detaii of the infomaticn about passive soakuxq .es not unifarn for all study subjects. Nevertheless. infornetion about eapOsure in the hane envirmoent %+as eitiier available or oould easily be estimat.d for everybody. Analysis of the data shoy+ed no increase in risk of lung cr+oer associat.d with ever hsvinq lived rith a seoker lat - 1.0. 956 CL: 0.64 -. 1.56). Controlling in the analysis for the tir'y'les aqe. sex ard study area, did not um+eil uy relacionship either. The et.tnber a! years lived with a soker .as not foisnd to affect the risk of lung ca+oer. Conversely, saae pasitive findinqs vrs'e descziled for e3qosuze to `x scrxise s smdeinq. A adrinq speuse .as fousd to be esrocisted with an odds ratio of 1.47 (not statistically sigsifiortt). 'ltte nunber of -elati~rstup cigarettes sndce' a ywk by the sMna shorwd a linear -16- vltli risk of Lv+g canoer deapite the fact that the ..*+nstes for ..ch level of exposure. iandivicti.lly, did not r.r-1i sta t.ista ca 1 s.ignificeroe. 'i'he highrst ard only siqni!iont effecz was reprxt.d for .omen 63 y.ars or older aarri.d to blue aolLar .+orlxrs .ho ae-e heevy srokers i280 or wore cigar:ttet/aeelc) . Oiffl3t S1S)E)IES. Oespite the fact srre studies were not intendsd specifiolly to look at the assaciat.Lm betr+ean passive sookinq and lung canoes lor at least this relationship wu rot the nain foau of their iaquiry). they provide inforu.tion to oonterd that such oorsrcrtion is feasible. Ztr study on lung canoer tsmds published by Garfinlosl (24) , as well as a studv by Rahet ard wynder (74) on the Ristopetholooic petterns of lung cancer in non-snokars , both fall into this eateqar.•. Similsrly. stviias linkinq passiva snuldiq with car.oers other than that of the lung can be helpful in delineatirq !he carcirnqenic potant.ial of sidestreaw sroke. 'lAe studies by Serailer and oolleaques Ili-]71 belap to this croup. Garfinkel t2t1 in `us analysis of the Mrrican Gneer Society study ocurpared the nactalitv rates for nrn-snakinq wren nsrrted to nor.-smukers vith t.`e sactality rate of rwr-vokifq vamn asrried to WK*ers. It .+s cicaret'es foind that w+en .hose hLLst,uds woked less than 20 per o+y had a 271 tugier 1Tq =ccr mrtali;:• rote ='an

uar.n Rarris,d tc nn-rrxers. T.x ratx was only lOS n.t7ter `or thase married to a mdker of aure thAn 20 c1?arettes per day. w1Yn confaadino variables .ere taken inta aesviait (aqe, ra..~, edumtion, residence and hus'y^^•' ooasastian). the e:erss risk was ]71 and ss, respectively. None of tfrse findinqs rere statiatically significant. lks.ever, the results have to be intasprned with great cautaor. beouse :hr study w never deaiqnad to olxain lsforsetion on passive srorc~, ftutJeraore. whatever smdcinq infors.tion was a<ailnbl[ carrespQ.ied only to the ejqasute at time of enrollment into the stucy. uabet and Wyrder (3Il assessed spoaue to SeCondary sroice in a subset of r+m-sioldtxl patients included in a large stuty of totacxv- related canceis. Out of 25 srle cases six repacted exposure to other peopls's ssdse at hane. as aspared to 5 ait of 25 oQ+trols. EYqvrea+ cf thes+e cases reportal esposure at wark versus 11 of the controls. tRis last differerce was Earely statistically siy.niticarx. Nvvq tne femsle qraup (n..51) aba:t the sass pcvportion of cases and cmtrols •ere e:,raed to o•.her peaple's ssdce at honr and wrkt conser,uentlf the association with l:+rq canosr w rot statistically siqnificant. Sandler, et. al. reported ttre rssuits of a case-crntrol study i:rml•xinq S18 cxncer patients (all sitee, except be.sal---+ell cancYr of the s'<in) and 516 controls. In three separate p,abliatiors the autlx=s assessed the associativ: bet.+een thesie canvxss and: 1) e:wosure in ct:ildnood and adult life to the ssocinq of all !ruxhold nrrb.is who s+miced (35). 21 early life exposure to parnnW sroki-q (36). and 3) eipasure in adil:2nod as a result of spr'we "s smicing (37). in the ''---st of t'rsa :;c jdis •s_io for G-cse ea~csec to ore a.:...r_r tzDze'CzOz MM -.le- in the hoauetnld w fourd to ce 1.5. If exposed to the smoke of tw indivi.luals, the obis ratio uxzZased to 2.3. . and to 2.8 if ex,-rned to the smoke ptaduced by three saokers. lhis trend was statistically significant (P < 0.01). tkn.ver, it should be noaed t.1at an i:rtease in risk w also uhserved for active ssokars .A:o had peen exposed to the ssdce of others in the housetnld. The oserall cancer risk for all study subjects (passive. as well as active, smokern) was reported to be elevated in both of twv circuTetanoes: vhen exposure to sewrSary ssNmd had ocursed duriiy childhood only /OR-1.61, anai .R:en the sagosure tooc plaae in adult life aslusively IOR*I.SI. tbr an individual esposed in both potiods of life, the odds ratio exatea:ed t+.o (OR•2.'). tbr the semrd repact. infomstion on ssokinq habits of both parents was obtained for aLmst all the cases (n-438) and almost all of the controls (n-470). For all carner sites the risk was inc'e.sed for all the cases (again. pessive.and active ssdcars ctnbined) if the father was a sroknr 951Q.: 1.1-2.0). Mstsirai ssWci.nq, AoweMe:. did not ses to inertise t)r ovarall cancer risk (Onr 1.1, 956Q.c 0.7-1.6). Siailarly, the risk tor lu+g cancer was irnsassed in those :xpoeed to t.1e father's smoke (0A-1.8, 9S9Q.: 0.5-6.6). !ut not to the not.her's lat-O.f. 9S%M: 0.4-2.1). hstly, wrt,en cancer risk was assessed in relation to the spouses snokinq habits, it w fourd that iz+dividuals warried to ssckers had 1.6 unes the risk of individuals sarrisd to non-sndvss (P<0.01l. 'Tis ditfererk_ could not be scsaunted for by dsfferer:ces in aqe.

-l?- racl. sex.' ed.xatuon. or occvpation batrees cases and corttro? 9. Wien site-speczfic cancer risk wes assessed, the auchors found that rxpcayad individuals (t-bse msrzied to sndcers) were at hiqher risk for cancer of the tsesst (OR-1.8), of the femele genital orqans IOR-1.81, and of the endoeiux systen (f1R-3.2). Based on a small nsber of cases (n-:2) it .as fouid that passive s+aokers showed 1.9 times the risk for l;nq cancer tn.n those not expased to secvndhwd snaxe. :his fi»ding did not reach statastioi siq--tificance. ltmre are mainly three oroblenatic features in Sandler's publiations that deserve some attentian. Ihe first relates to the .ay in which the resul;s have bem presrrted. Ln many instances the risk ror anoer associeud with passive smoking is prasented for the whole study qtoup, ca'Qrisinr non-sohers as well as ssokers. Althouqh the latter group car be eonsidesed passive saokers of their own sec*xidary sroke - as well as of t.he anoke that others preduce - their inclusion seeinapprapiate, since the ouenAielming effects due to direct smdcing wuld tend to distort the effect due exclusively to passive smokinq. Scae tables, haxver, do seJce the distinction betwe-s the two qrojpe, and those are t1r- ones that are noet infornetive. Secadly. confoundioq variables that nay explain the otserved associations were not taken into acoant. Sone of these vwiabies are expnsure to alcohol, to ots.vpstional harards, to drvqs, and to sacuai and reproductive belrvior variables. 'Rtirdly ald lastly, the 'act t.'at an association .as fou+d betsrs assive mokinq and cancs: s not previously deseibed as tobscco-related (breast. ~n!r.__.e svs_emi raises .ae suspicirn. rhi3 fi_idic~y r.ay be a I t',,~ .:a~'~R- -20-. conserfienee of the inability to control for i.rnfouudess, as mertiored amve. In addition to the epidemiologic studies.. labaratory im+cst:cptions, along with elinial and pathological observatirns, have sipportad the hypothesis that lung canotr and passive smdc_iq udeed t.ay be associated. 'IIhe enet relevant literature in these areds is discussed in the follorrl+g paraqraptr. 4 IABOPAICRY INVESPIGA'ITQtS Both otaervational ard experisnnrtal lamratory stvdies tuve produeed valuable informntion on the biological plausability of the health effects of passive sseking. The mcst iaportsnt findings are related to the identifiation of carrinoqenic substancss in sides-.rers snoke and the ability of sow sidestreem smoke caipow+ts to enter the organism of the passive ssoker (38-45). Cigarette smoke contain+ over one thoLLwd substav+cti (46). Meny of thee. including carcinogens, are present in even higher aonoentraticns in sidestream sn,ke than mairutreem smoke. Ebr ins•liace. Dincthylrutrvsanune, a nctent carcincqen in an.iaais, is i2 t:nes mxe wncrntratej in sidestrd-&m smoke (47-48) 6nroenann, et. a1. (49) have estimnted that a non-. snoker wfw spends one hour in a ssoiry roon nay i.'Jiale nitrosaminrs in quantities equivalent to snakinq abovt 15 ciqarettes. Benzo(a)pirMe. arot.her !oioun ca-einoqen in nrr and aruauis. is present 3.4 tises as *~:C ;-,. s_'cstr•wn sioke. EU_^hemore, rne agarett- Tay release up

-21- . to l0u aq 'of L-ta *_ subatanoe i.'vto the air (46) Jut hov much expos+re to these subetarxVS docs the regular nrn- slokioq cers..r reslly eaixYienoe? This has berJ, eery difficcvlt to qMn•ify. 'O+e ocnrxn:rations ef too.ccv couponents in :`* envL-ammt depends on nunerous factcrs. The aoant of tobcco coRs!r~d, rl+ volume of anblent air. the sise of the roos whrre wokinq takes place. ud L'w type anC aseumt of ve+tilatlon in an eeclosad space are sane oi the mur1 variables that alfect sadcs i+take. 1he iroxi''oty of the passive smo4er to the souroe of sfmke and hislhRS pattern of inhalatim sey also affect 'actual esqxxure (48). Similarly. tne sndcing patterns of the itdividuls who do the srak,ing need to be emsldered. tbr instance. the variatiotf in puff frequency. puff duratirn and puff volune are knorn to affect both the chesical r.apnitian of sidestst.m -woke and its Lsiological activity (50-54). EVen the smdcer's hrlth status saes to affect the wey in rhich sidest:esm smoke is produoed (55). In fact, it is the sidestream swke that has been sodified by the above and other factors tSat eanstitute the real .xpoaee to the passive smoker. ScRe authors prefer to refer to this as envircmmental toasceo smdce rather s:3est_eam snoka. dWl Desoite tlrae difficvlties, srne studies have attewted to quantify actual eaposure ty swsisinq totacs=° aetabolites in tuty products. Sena thyocianate and cvtinine are !ourd in the blood, uruw, serue and saliva of the passive ssflker (45, 56-60). Alt1`.ou9h aycrianate auy result !ran other acurxe -- i.e. leafy veweables - -zz- cntuune is a specific nntker for exposuxe to totncos snnke and tharefore its areserre in the oon-wdu~ss is takerr as evidence of passlve innalation. In onr experimrnt a regesrclier was also able to identify .utaqnic suD.tanoes ii the urine of subjec--3 exGnsed to smke in a poorly ventilated room (40). Studies aised at determilinq the specific deposition of envzrarnrntal smoks oomnents in the lung tissues are very scanty. In theory, the biolaqiol features of the anst®y, aand physloioqy of the himan respiratocy systas., and the physial and cteoucal dkracteristacs of sidestzesa suake, suqqest th.t surh deposition takes place. 1be extent of it. hovever. has been a satter of debate. Dl`fesmt auttnrs have reQorted high levels of ssake brvndti.al depoeitim ranging froa 114 (61-621 to 80s (63). rrnuereely. 11ep.cs and Inwry (64) think that .oet pssaive smokers arm exposad to a edniasil asount of asda, perh.ps rn s~ace ttnn the quivalent of csle cigarette per day. It has been suqqested that sica an exposure eonveys a negliqible risk (65). With respect to this, it should be notad that no treshold level of wqause for the develment of lunq csnaer has been established. 7trrefocv, any level of expasure -- includinq the tow level attained by pessive smokinq - should be c+QUidertl pntantully able to elicit a oareirogeruc response Another ssies of studies have tieen aiaed to detst the effects of passive suoking in the respiratory !unction of passive ssokers. '['Us ia prrticularly relevant sir.oe sa+e avt.hxs think that an

i -23- u,P..=ad :espi s.ory tunctton is in itself a risk fact:.r for iung caneer (66-69). In Ftanct. lautf'nan 1681 at. al. cvwnred spi:amtzic neasureventa oetwea~ tso types of nrn-s4okers, exposed and non-expased to seovdary snd~e. 'R+ey found that both nrs and .cmen mar_ic.' co soticers of >• 10 g of tobsocn a day had a significantly lower forced mi`-exmiratory :low rate (t'L' 2S-7S) tnan t'nx onctied to non- smuce:s. In addition, w;an in the axpceed g:wp shr%wed a decre3se in forced eouatery volume in am seccrid ;''V! 1). Zhe study ca..''fu::y ca+t.-olled fs contotimdiJS7 variables sum as age, sctaal class, echucation, family size a.d air pollution. White arz' F`-+oeb (691 eQ+ducted a study in whidi the effects of passive sndcyp in •hw wclcplaoa wrre assessed. It :as fourd cJat the decrease in foecad mid-eviratocy and end expirawry (FEF 25-73 and iL' 75-95) amrng passive seokers .as ocnperable to the decrease obsernzd in light swke--s. The effects of passive snoking an the puLmna.-y hnction of cnildren as eeasured by spizanetric qieasuremr;ts was Mentioned in a prroeeding section (13) rthey have been confi.c?md reoc;tly in a study . bv i7rn an.! Li cvduRed in Shanghai (70) D:.`_er~~t ca:cinogenic mecheatis+ms have A+c; postulated `ar &T+dce bot.h neinstream and envi.rormntal. 4s•se aectanisro are rhoixght to esplain the develoPnent of lunq caaarl mtiq passive s+rokers even i•. as some belie+e, the aroke dose cmnferred by oes.sive smdcir.g is lcv. Such nethanisM are also believed to be rela:ed to tAe dif'_ere.11: .".istolooic per.erns obse_tied in smkers ar.A na;-mokers. .=ccord:nq co -2~- t_tie t--st hypothes:s, ca_^noyens preseit in envir.raen al smoke rauld hive the ability to producY adeioc;rciroomu, the type of canczr more camon aaaq nrn--wmlcers. Sirce volatile coaxxxunis are more :ik,ely tc be abaorbed by the pessive smaicer. it is precisely in tlnse c.naounds where caro.roqenic propertie3 wuld reside. f4rtheavre. sirre volatile carciuszens would be able to rrach the distal ends of the brtndual tzea. one would exp.et to find more twors in the ,.eriphery of the luar,s of tm-sokuug cases than in soiqng cases. Mynder and Cmdoan (671 suqgest tht, itdmed, this anatunital pr-cerence is pre>snt ancrq non-s+nolarrs. 'Rie sec.vid (.ost widely acupt,ed) hypatlrais pcvposes that envirornental totsooo - lce and ralnst--eam smke have eactly tne same carcihaqenic p~copertSes, despite the loroiwn diffeseiwes in their pf+ysimcheaical characteristics. ttrder this hypothesis, the esposure to ern.ironne~tal aroke is mnsidnred to be aquivalent to low levels cf eaptauri to sauinstseo.n ssok.. 'Ihu faat that no treshold has been establis.led for rhe carcinogenic effect of eaintr.re :sdce (tiiet is. that any level of expaaure cviv.ys a risk aboue Ch.t of no exqr.uore) stpporis this hypothesis. Plnther suppott is provided ay the observation that epidernoid and ~ll acll carcirnme (consider.d by mary to be the namors nore strongty related to active smoking) have been fo;rio asaocar.ed with r)Qoeure to passive sndcing (30.33).

5 AA'pAAL MPEltlf'fMS. -15- Testinq th.e hypothesis that passive snnking and lunq cancer are related by searL+ of aru,el P-XPeriaents involves similar Cifficulties aaciW to the study of the carcifcgenic effec~a of direct smking. The search for a suitahle aruaal s+cdel haa taken se:ny yea.-s, and 5hile resasrCie.-s are able to reproducs many of the respiratory r+stem tumrs. no one wuld claL+ed to have found an ideal sodel. 1b+ver, even vit.hout interdinq to do so, seany of the experinents aiaed at soudinq the effects of direct smokinq sey indeed have been testi;^y the effeets produoed by passive sedc3nq. Various species of lahoratory anisals (wioe, rats, syrian haneters) have been eqerisantally expased to srcice•-fil3ed arnirarsner+ts (72-75) Qne such stuiy (72) nw eonsidered a classic in the field, has shorn that rahhits expoeed to em+iroroental ssdce s.y develop traeiwobrcrlchial epithelial metaplasia and dysplasiL. This study, along with the others ahich followed it, rere interpreted with nLkch caution since they did nct quite sisulated the ptmamenon they were intended to study (i.e. the effects of direct ssdcing). Diffeaent but equally ir.pcrtsnt caveats should be taken into consideration whtn these studies are used to draw conclusions abArt exposure to pasaiae sndcnq. For instanoe, we eannot be sure tiut exposure to snoloe in e closed c!&nbes actually sinulatey pessivs andk:nq, nor that t:.e wuta.ry and physioloay of tne respiratory tract ef the experimeraal animals wuld :espa-d to secrniary srk.. i~ the samn_ s,ey as huaen'Y. ~bre re:ently, soFnistirated eou:rnen*. that ?rodur_s and trios sides;re.am smoke has -26- been developed to simulate eaq~e to wi,ole sidce or to its gas phas4 only (76). S1ich devices include wqqosure chaobers for rodents. oell cultures or isolated perf-use7 lunqs. as .ell as aect.a+usffi to sanipulate sm*e volume, dilution, and other variables to be studied. Ttiese devioes are not yet part of the stanizrd equLpment of spenilined labora:nries. but it Ls likely that cheir introduc-,ion will prodtce a wealth of knovlelge reqardinq tl.v effeets of sadest.-ms suoka. [d.stly, aany studies have presanted eeidenee of t5e mrcinoqRnic propertles of tobecco oteporrrrcs .hen adeinistirad ttuough rattes other tnan the respiratory t-act. Many articles on the effects of subcutarwous in7ections of tohecw aomporsxft.s (76-77),, alonq with artic!es on the effects of rtbb,inq ard skin prirtting with tobcao corpauds (76).. agree that there is little rea.wi to doubt the caranoqenic prvper-ties of toCsooo sscke. 3UIQNG 11ND HISEC'1MIC °YPE OF IUNR" C11ltCfR An oeaeriation that has created a grst deal of interest amrng rea.archers euntzrns differences fouid between lux1 cencer cases in snokers acd lunq cancer cases in noRsmicers. These differenaes pe-cai.-' ira:nly to tw factors, sex and histologic type. The relevant literature in t:us sub)ect is disaused in the follovinq paraqrapRs. Mvnl .amen with lunq caneer there is a higher percentage of wtisncke_rs than amxrl eale cases. In upper !Mw York State - as

i -27- pcelinunury infcrAStioo for t1as study was bainq coliec:eri - r+e :ourd that nmsmdcuq .Rm.n owprised 9.01 of all few.le ase3. Whereas nonmdcinq sen reprEsented enly 2..01 of all sale cases. Using data published by Oreenber9 at. al. (781, 1 estiswted similar figures. In their series of cases. 10.71 of the wen ..re nonemkrrs versus 3.0% of the m3n. Fras data provided by ltaost and 6lyrrSer (34) the estimted carresprndinq figures are 11.91 and 1.96%. EYnally. Garfinkel et. al. reported that 126 of the fe.nale lung oneer cases sczeend for parttcipation :n theis stady 1301 wete norrsndcers. It has bsen suqgestad - since ferer wRn than men sovke -- that a lazqer propoction of females art left sueepti.ble to the effects uf other a4rcinvfens (includinq seomdrrd s9e1u1 than to direct ciqarecta sndce. 'f1+ese wrcitog.ns wuld be responsible for Lhe "exarss' nurbar of nmm=*irq female cases. 'it+e sffects of these carcina3ens in srn wuld be mask+ed by tne overwhelming effect of direct ciqarette amdce. Histoloqic type has also been reeognu:ed to be asso:iated ~Itti smdcing habits. 'Iha nature of this association, hoeev+_r, remai.•u controversial. As early as 1957, Do11 et. al. (79) had proposed a ^Case-resprxtse' reiationship by sho+ir.g an usociation oe_..aen the aaoimt of _cteeoo ssoked and the d..mlop.eiK of e;,Ldrsfmu'., large cell and sroll cell curitor.. Qtrer listoloqie types were fxed to be eit.'te: w.:ela--ed, or only sltqhtly related, to smuno. A few irears later, !c:..ybecq wblishrd similar observations (a7). He fourd an assx:ncir. Y_-reen seckirq and epidersoid, large c--1: arc' Tmsll ;a:; c3: clnrna'.i. -=e- Similarly, he re3ected an assoc!Acion be~ ween snok.inq and bot..h admocatrinrna and hres+ctuoa1veolar earei-nrn. Etvs then on. ro autt.ar has fully ay-.eed with any other, pcexntinq results that are often fradrly cm+**ad;ctory. aeiss, et. al. (61) in a prospec..tive study of 6,163 srn, added adef»carci.rams to the list of histologic types related to s+ok:.nq. Yesrr_r 1821 crnfilned the association betreen s.okisy and soell crll cacaraia. but rrn with other oell types. IWerbsrh fousd snok:ny habits to be equally distributed in a11 eell types. tAeretcre daryinq any association betrieen histology and tab.ooo sndoe (83) 4inoent cvicludes that edenocarcinmis and btrnchioalveolar areirnom are irrvers4ly xlated to snoking (84). Md. more laaosret.ly, tw iadeoendsrtt studies have repart.d that all call typa, incluainq adriaarcinmas, are related to cigarette ssdcieq -tlS). ' The difficulties in intespcetinq the results of these studies are many. Two in particular, however, de.ervr_ to be eentioned. tZrit, there is the problemstic characteristia of the study qro<Qs. 7tey crnprise such heteroqenous - and selected - populations that "sible smple bias is difficult to ignore. Secvndly, the intecvention of so many pat2troloqists in 1» reviewing of the slides brirqs to mind the likelihood of unreliability of diaq<rxes. It should also be added that the studies cwer an extr_.rded period, t!r aud fifties to 19N4. During this ptriod the ability to assiqn lung :ancer a.ses tz) sQeeific lustoloqic types has ctanqed, as can be aisessec.' by t.tie ;.ar.-rasinq nwdnr of caxc d.iaqnosed as

uc'uif_ren,tldte.i ==11 types. sex. A IR'L'7 r.lk'.rC aSSof'1ar_Ofl _s fOUId betNef'l: hLStO!Oq1C :ype d71d "t studies sqrec that epidermid cfl^urma is a=e f:equent ia nw?es than fewele3. 'f!r prcportion of male cases classified as epld,?r.=d carr,r,r-3 ranqes lmrr"en 781 ncxl 64i. fbr fmmiee, the values are betveen 21% ard 551. 'ltie other twtab:e dlfernr_ occas a.ncr.q adeioarclnrnas: this tisor is earh more =wrn amon9 w+en tMr: nm. :)p to 521 of all Iw.q cs.onr tases in femsles have been rep<sted to bel011g to t116s! QrP.pr Mhereast 721 is t1f- lll9hl3t aQl"r1q Ml1 to be prey¢lted in :e[.lnt studies. 'Itte differafrr u1 ssaan9 habits bets.een .en ard w~an aug9~ dat suc.h hacits mY be zesPoKiD1t for the differsme in h,istoloqic types aarY+q the sexas. Ilnfortun.tely. this l+as not been ProperlY assessad in any scady. Most repects deal with the sEx-hiscaloqic typt, or the stok.in4-hinologic type association. 'fie thsee-~aY iiReraction, sec-aokiig-histoloqic typa., remins to be studied. o~~,~&CCzoz Qi71PCFR 1%D lEZtiL'S In Aoril of 1982 the Nev York State Deoartaant of Health initiated a larqe study of the tpidenioloyy of lwq owxt. Neu york State, with appraocLertely 10.000 csses of luty carner reported every year to its tumr registry, prwided an eaaoeilent set-._inq for a study of this nature. She project aas iMerderd to Ee caapcettinaive, that is,, to include both snokinq ard nort-snokinq liny onaer cases. Study of the qrroup of srdtlnq cases was limited to the clinical ard patholoqiol diaracter!stics of liaq can©er, whereas the assessnant of e:qPoaure to ptssive smokinq arxi other relevant envirvrawrtal variables wre vx main study ;rariables in the non-smking y:oV. :'u study design also calied for a population based cax,mettfied control qraip (this qruvp will be described later in full detaill. it is only ches. :w latter qs.Ape -- nm-smokinq mses and their mstc'rti cantrols - correspordent Uat will be used to assess the effect of erprzure to passive snokinq 3n the risks of l:m cancer. The -30-

UM mechcdoloqic aporca~h .ised cc :h_1s err.' is t.'r one-to^car mstctrel case cantol studi. Lt oay be pertinent to nmtian thit the ori9ir-a1 desiqn of the srsly s9ecified that,, in ur3er tc dinaern sore pre~_isely the effects of passive soaicin7 on the risk of dewe;opisy lung cancer, all cases included .ould be never smoicers. 'iwo tAin9s vete clear early on. 1a.evci: firsc, that sale cases wrse rarely never sorrkers (only 2.0% of tlxnn initially screersd for inclusion in the study reported to have never sao.t,edl, and sec.-ond, CMt a suf:icient nunter of these cases onuld nnt pcasibly be assanbled in a ressonnble per:od of tise so as to sacisfy the sarple size requle+rents. 'lfiierefore, the eleqibility critusia for cases ims extaded to include thuse rho had stopQmd snoking at least 10 yean before diagnosis. It shoPVd then be clear thet .ren ++e refer in this rrport to the qroup of non-siokiny cases, we are referring ndt on!y to never sROkers but also ta :orrtnr snokers. Tfie :.rpl.icatioru of the dedsion to include for7ner sr:rakess in the study 9roup are discussed in the later section of this C-'tispLrs that deeis with sethods of analysis, as well as in --iapter Fbur. OBJI.~`TIUES . The msin ourpose of this srud., w to umiore l+e relationship aetves aassive sm*unq and lutiq car+oer in nrn-snokxra, uanq a msE- c5.t:c1 apFraad:. 1Tas tias :`onp by anslvrLn5 ?.sta colle^_ed Ln 439 cases Of lung eanaer and 439 mecclmd 'healthy' populatinn canc.rols. the specific question addressed in the aralysis rere: 1) Is there a;isk for lung cancnr as..,r;,ted „ith ty spouses • su*.tirx'i hahi's? 21 Is tirie a risk for lung wnosr associac.d with snkiny habits of asnbars of the tousennld linclusave of soausel? 3) Is axq:o.ure oo Punivg adunq in the %dxkplaoe associited vith a hioer risk of hs+q canoer? 4) Is passive aokinq in social aitust3ons s.aciatsd with a higher risk of lung onoes? - 51 In the .v.nt thtc ths ans.cr to these yuestion is affirmet;ve, does the aa.qc9at,ian p.rsist aftar oontrollinq for c4nfo.rdinq variables? el Eyn a dose-response betw.en passive sNokiny ard lung carc,~r be sha+n in this datat 7) Are there ditference in risk for aien and wmen? !1 Ate there differwuis in risk tor never smokars and fcr focter sed:ersl 91 Is t.here a Riqtie: risk aasoci.ted vich a Specific lwg cancrr hictoloqic cype]

JM -33- 2 SAt9n€ S.1E. It was estisated that a sanple size o: 450 cases (225 males and 225 fenales). toryatMr with 450 mntrols. would be risaassary to datact • relative risk of the order reparted by 8i-'syoa 25) and Tric'npulous, ec. a1. (27). Pnelisutiary inforrati,on c=llec.ed in the study area showed that aporaocLrately 1,200 fan.le Iung Car.?er cases were diagnosed in the stuiy area every year. ard ttit 9.09 `ere a~trng never saoke.<a. liius. it urs escisatad that (allovinq for a 15% refusal rst•) the 225 faswle mses muld be assemble+d in a ceriod of about 30 c+a+ths. roughly the tzse available to collect the data for rt.A study. Since only 2.0s of san .are never ssekers, garha:ing the sais rns*er of male cases vould have taken axA ssze tim t,tian the avsilable Rricd allo+ed. M indicated aboue, it .es sueh a realization that prrnQ_ed the d.Cisian to include .:aseY chat had stoFped sndunq at least 10 years cedc. The final saeple size fell slightly st+ort of the petAxwed ruar+.x:. it was caprised of the 439 Casas and their 439 aetct+ed mrttrols, for ;Ahi:h infocaation w carplete. MKr:g all eligible cases t.4: respcrue rrte -as :lose to 509. qtbtrac-_inq an additianal 4.01 fo: ohyaicians' refusal to let their ?etier.ts parti:.ipete, ttk to~al resconse rate can be esti~ted to be 761. ~TVzSCMz -34- 3 IGCRiTQ7 OF T.l"c SMMY. A d~eck of the data acctsylated by the Nev York State 'risor Regiatry shnued that about e0s of cases of lung carosr ocnu amax3 reside+ics of the eight Standiird Metzapolitan St~ial Areas of Lpstate Nev York. lhis locale oos4cises 23 caar-ies. with apQroaia.tely 125 diagnostic and/or tr.stment tarilitie.. It w decided that the reaouats aaaiiable for the study would be 4sed sze efficiently -- ard the loqistio of the field .ak greatly siaplified - by liaiit.ing Ow reudy to that q.oqraphic anea. 4 MD18LLShT47l oP '1n OlS€S REft7ttTf[: SY5'I9T1. 'lftie first stap of the pmfect inuolv.d iattirr; up a repor_inq system in a11 participating hospitals. lhe field staff, oraVris.d of six e)qerienoad Aesearcfi Asaist-ants, .as in chatyt of establishuyg the necessary cbrttacts within the irot.itvtioro. in e.ch hospital tr+e Medical Records pefartmr:t. the Patholoqy Deo.rcnent, ard the Ytaor Registry reported to our field staff all clinically and histologlully 3~,'oeed 1un/ ~Jnocr cases, either via telephone call or duruq `r.e field sca;1 regular visits to the inos,utal. Cooperation trva all par_LCipsting hospitals .as es.llent, with the larqnt facilities ceportinq as trequently as orre or tvice a.eek, and the .aller ones reporti.-+q at least onoe a srnth. t7u t+d. York State Cancer Asqist_-y +is als!+ screened ;),-r:o'r.ically to asoe..^a.i.n casos t`vc sugt,-- have be-m

-35- oussed by.;Je hotpita!-oased remrting syszem. Specal enphasis as put by the field staf` on r~e tamelinass of the reporting. suKae it was neorsssry for ur_ qupuses of tht study to obta.in personel iotervyeas with the cases before they died or beone too ill to perticipste. The aeen tise elapsed Eetwea: diaqnosiY an3 reportity tias estarated to be 45 days, with an additional 102 days transpired bejor• the COfduC~ of the lAterVley. 'LTu reMrtinq of a case wss followpd by the securi<g of the snayciM r.uvory stated in the pstients' hoepi.ral medicel reoords. All cases reuuted as nevOr smoke_-s- focter ssokero oc of un)o"M srok.inq statvs wnre oontacted by t.lephone ard ttrs.r smdciJx) history c<s%Ei.+ned. trnly those canfitsed never aeokers. or ex-sndrers for at lwsc 10 ywarsr were oonsiderd for inclusion in the study. A11 the eleqibility criteria for cases are list.d belov. S EIMIBLLLTY Qt=RIA fDR t71SES. 1) The patie+t wnt be a resident of the 23 casrty study area 1e 3611 of Upstate Ner York). 2) 'Rie patient had to be between 20 and 60 years of +qe. 3) 'Itie patient had to be a never sroker 1 ssrked <. 100 c.garettes ?n a 1=fetimel or an ex-sndker Ismoked <- 100 eiyarettes in the last 10 years ). ECVzSCCzoz to -36- 4) ltie pa:ivit should have a clinical or lust,oloqical diaqrra+is of pcimary haiq canoer nad4 bet.fen July 1. 1992 ard re++++or 31. 1984. 5) :7ie dis4iarqe di.qnos;s tad to be ctnfirned by the re- erAmination of tte pattmlagy slides by an expert reviewer (see Nisto logy Review in Vhis Captez ). Cam-saly, the exciusim Qitetia aere as follo+s. 1) Any petient initially reported as anever or former smokrr, who w later foaud not n t=ply with the definition of never and (ormt• sndrer us•ed in r.his study. 0 2) Any petient with a diactarye diagnosis of Qrinary lung cuxxr aho, upon the re-examination of the Etistoloqic speciaens, rws assiqrti,d a different dieqnasis. 3) My patient not fulfilling the nquirennts of age, residency, or date of dieRnosis. 4) sny patient for vnan an L-xftvidual or a physician's cons+ant .as not qrantai.

-37- 6 Sc^.-.OmON OF GWMIS. Li otder to provide a eaW-LSm 9raP. each case i¢ividually satched to a popllation cantrol. R1e source for weis such conuols W43 tne State oepertar-rt:t of Mota vhiclss files. 1h s sou:ce wes Lnfornatzon necesary to perfoca the aru:hu+9. Additionally. .eLhod +s rew3ed as ;ess tue-cw=in9 and mare eco.onucal coruioered apprvprute since it pcwiaed moct . of the ttu s trwn other possibie oethods for the selecZiir of m+trols, s:eh as random diq,-t dz.luq. For each case, six potential emtsols +rere select.d a+ t1e o.sis cf aqe, sex, ard cosity of residenoe. llpan a teleohax interview. the fisst pxantiel control .R+o .es foux' also to rsstc-ti the case on tJre oasis of ssdunq ststu+ - ard who, in ad<iitian, aqreed -38- An additional mstdunq variable was crnsidered at the ume of corductsnq the :nterview. IL cvnerns the oattnr of whether t:,e questicnna.ire was resporded to by the case himselffherself or, an tne other hand., by a wrrogate Eespori3rnt. the sss3ting on type cf intennew is better esplained in the sectirn of MM 07~a; vhich appears rsect on in this Quapter. On the ave.raqe,, tw pot.nt.ial controls had to be called until one aatched the case an sedunq history aid was willing to participate in the studv. to pacticipete fully - ras included in the study. In ssn"ary, the eliqibilaty criteria foc oaetrols w as folio+s. MPl1 COLII~PIQd. Gnoe the eleqibility of cases ard oontrols was determined, the 11 The oc*itrol had to be of the sase age 1* - S years) as field staff arranged a face to face intesviev./hich took apormcimtely one nour to oonduc^t in the patient "s hone. Al l lsfornetion was the corresFc.din9 case. crollected using a pre-crded Swrostiomaira. The quest.iarmire .es prc- h tested for a 3 month-period usinq patients dieqrneed prior to tne 2) e 'R* control had to be of the ssve sex as t study period. The sections of the questionnaire that oontain the 1) The controi had to be a resident of the same itms used for analysis in the prexent study ars foLrd in Appendix A. cv-Mty as the oontrol. ~ As wuld be expected, ieoat of ;Jese questions are ornczrrx.,d with che Prasu:e+ent of ea;oaure to passive s+dcinq: 5ut infors'tim on soc3al, 4; '1`r. cu+uol had to have the sa+e saoK:ng histMY as - er never xrokers or i*1 d t o h h drAeqraFruc and nediml varzahles %es also souqht. Both cases ard , e e a o ose. TTUt is, iro[. c5ntrols ..ere interviewed eamctly in the sase fashian. and exrept for ex-vrokers for st lesst tO years. L11e 1teTS L'1 t1M1 Qllestlcf4Nlre referring to the :LL'llcal aspects of fUZgE Czo%~O hMMM . ",,e

-.39- t±e c.irrenr meiical condition., they tcth ans,.c--red the sam rnmorr and tvpe of questlanS• ti,ci t.'e fie2 stafC faud that a case hed died or had beWane too ill to percicip'1, he/s~e would ask to Oonduct the irrterview w_a a next of ?cin. preferably .at.': the s,,.^ouse. but children ard ot.her relatives were aiso considesed aproplsriete surn-,qate respordents. rhw-n a sL'*tyatn respondent had to be used :or a cax, a surrogate respcndent was also used for the ocntrol. For instance. b+hen the spouse of a case was interviewed in lieu of the ase. we would seek to :r.terview the spouae of the person selected as a control, even if the latter was available ard .rillinq to be interviewed. In this respect .e shoJid then note cMt cases arra controls were artched in addiuca to SEX .!GE , BMOCi}G HL4lVRY ard RF5IDfh1E. an t;h TYPE Cf INMVIDi conducted. e H1sIO1= nFVIfM. M independent review of the pathology sF.edmens initially sra.nined to sake the diagnosis of lua+o onoer s.mmeA urportant for _w reasons. Firnt, to mske sure that all :.a-+es were con::rned or-„rar)• ILM cancer cases. ard seaord, to prwide a uniform c_^_te: La of diaqncss. 'lt.e proordLre fs suc!'. review was as follovs. '.`,e f:elj* ctaff .cild r_rt.a,.-t tne p.raoroqy departarnt of tne -10- diaqnosing hmp.tals and wu:d requrst the slides or tissue blodcs for the patient :n ques-son. In all b:t five patients sucn reqtwst was suc=sesful. Tfie spe^__nrns .ouid thm be sent to a Patholoqi.st at Marorial Stoane Kettering Cancer Center in New ycrk City, who wuld review ttnn 5lindly with rogard to smokinq his~.ory, the initial hospital diaqnosis, ard other risk factors. 'L5e reuiev pathologist assiqned each caae a histological diaqar.is .czordrng to tne f.orld Heaith Ciganizaurn International P.istoioqical Clasaifiation of '.Umurs (86). 'ltie review diagnosis was then c3npared to the initial hospatal diumpneis. If there was disaqreement brttieen i+-,tiil and tiview diaqrasis, a secard pathologist fr® the same institution .ouid reviev the slidks ard assiqn the final diagnosis. Even tJnuqh a detailed histological classification was pcwilel by the pathologists, 'or the purpo.es of these study we have used a ct+arse classification that divides our cases i-iro the Eollovinq histologic types: Epidernoid or squamous cell carcinon., 9n.11 cell carcino:a, Aderrxarcinare, Uuye cell carcinons, ard Others. 9 ©6OSUtE vARIAg1FS. :Te m:in hypvthesis of the study was conoerned vith the effects of passive xeacir.q on ltmq cancer risk. In order to assess the maqnitude of such an ansoc•aticn. we needed to have both actvrate sessursnm+ta of wqrosu:e to the aapected risk factor and accurate 71eararrrent of the diW+e status. 1A>wr the latter was ad.ievsi nas

-a:- Seen e.+plaued in the previon s..wtion daaliaq with his:ao1oqic :r,r.ew. I:I this secion it will be e>plained hoy the ;xposure to t.he risk :accor wes deteatlmd. . 9.1 9P7U4F 9MNG W.BITS. :t.ams S9 a 73 in the (rxsti- (1pp'wdu A) inTaLred aROut tne snui_nq tuo:ts of up to four spouses per study su5ject. That informati7+ %es used to a:a.tt the follwinq srauumr+ts of espoaae: 1)MLWe: of ciaarettes smokni ±ft Pt the s ls . 3fie nuroer reported by the respodent was talaa.i dizectly from the yirsu^^^U+-e if the case orMrvl tiad been sarrire anly orne. or if only one of hisR+ers spaue(sl hsd saokai. Zf eort than o,re .arria9e to  savker w reported. '!u exposure w estimsted to be the mesn rsstser of ciSarectes/day smdced by as nny s*uea as repnrted (esxinai 4). 21PteErs of Y*A_s the „pouse snovad. This fiqtise •+sa taken directly Eram the reprx'ted eamber of y.+ars tee first spou+e sroked witle narried ark' living togerhRr with the study subject. when apnlicable - that is. in the caee of s,ore than ane esuria9e to a aedke.r - the nater of years soolosd by sudcecP+ent saaues %as s:nply added to the first figure. J1'lbal nyrb.r of :i4arettes s~dced saoe drino e+s_=:ed life. th:s variable vs calculated by Tsl:yplinq the nurbe: of cigarettes sAokad per day by 365 to otra:n ^^ estuste of the xsraer of d?arrttes ssok.d in a ye.src; t~m, iruitipl,•inq tnat figure b/ the mrt,er of vaars the sooise w rep<rted tc nave TOk!d in -hR CDµrSe Of t!Y *lrtim'jR . 111lfl IIIC.re :hi'1 ~:nE >Po.ue ws :epor-ea as a sndker. the total married-life ntmber of -42- cigarettes tor each .,ould be adle+d into a yingle sesaueRnt of 9.2 DXP061RE IV PASSIVE SrOKM IN T!E BQlSEiO[D. 2his variable w created with the info®ation esllect.d in the itees mar(aed with an asterisk in prge 170 of Appendix A. Infocnetzon ahout smoking in the housdold ..s collect.d for the 12 so.t reoent residelnes in whidh the study subjecYs repart.d to have lived. lt w caiculated as follars: 'ihe nsioer of years Liv.d in eael~ residerre yas ssdt,iplied by the nuber of seoksrs in that particular r.udenae (znclusiMe of the spm u.1. • and added over as stic-Ty residetwes reported to have had smokers lwziinsn of 12). 'fie rasultinq figure ..s •x*+reasaid as the nuaber of psrsoNyears of e>posure expesienosd by the study subjems. It has to be enp(rsis.d tAst theas units of exposure s.tiould not be mnfused with the conventiorrl seaning usually atcaded to tlien in epideaioloqic re.aarch. In this case these units are used to sumarize the esposure of e.ch individu.l aooo:dusg to irnaroity (nurtar of sudcers in the hondnld) aed duration (rnsnber of years livinq vit.h snokers in the houselnldl of sueh aqosuse. TAey do not indicate the aollectiee experienee of a group of individuals over a oar-t,a.in ceriod of taae.

9. J PIl55NC 4KlICING IN 1qE FtORAP1ACE. The info-vetion collected t.hzvuqh the .teis merAed with an asterisk ir. paoes 168 aad l09 of A2pehiix awas the hasis an wnich to esuaace exFosure irn the wec!plaot. This variable .es eeated in Pxactlv ;re saw mervner i*s irusehold rxposure. :'fat is. the nurtxr cf years the stady suojx: wrked in a psrtacvlar place rs sultiplied by the nurber of smoksrs who wrked in the same roon or arns. This w.s done for the 12 .ost receK jobs ard added over to obtain a sinqle reax:reaer.t of esposure. ?fo units of exposise,.as vell as hoLLsatnld esposure, are espressed in person/yaRnr ard their 1Merpretatioa shouid also be sinulars a suanar; of ezpawre for earh stdy subject to sm,>lu•is in the wr!plxe over otrtain period of tiae. a -..- ;ccm zero to a Tsuman value of 96. This index is obcalned by addinq the infocaation mc'.ed in item 186-200. lbr further details on the aeatim of the udex of expoeure the reader asnt refer to the Jpprenl:x. There, both the list of t.he activities Sncluded .nd the .ey in whieh the information w entered 'u self-epLu!atcry. 10 PU.FJfIA11.Y Q?ffallD22t1 VA1tU+8=. InspecZirn of the data shored that cases and ornt_-ols were different vith zed+szd to vsriables wAiich :ae Eeiieaed to affec-- luaiq caners riak. It .ws decided that these should be eonsidesed in the analysis as potential cafounders. Me were awre, howeuer, that the inforastion available .s not of uruform quality acoss variables. wieceas inforas.tioo on sociodemoqrsphic variables w likely to be highly reliable, the infowation pertin;nq to ctemicai ard/or 9.4 PASSIVE 9HOIEI.V('i IN 90CTAL CIAC.M-rAt=S. occupetional exposures - all obta+ne+ on a self-zepocted ba.sis -.as not deened so. Ttws. .e carried out the ad juutmeit for onnfousiitq Paqe No 173 in Appwdix A sho+s the iteas asked during the variables in tw steps. 'fie first. .Rticfi we call •partiel' intervin+ in order to estinate exposure to s.cvxiary smake in social adlusonvst, incluled only those variables for which we had 'hazd' ci.rcurotarroe+. These itess ware intended tc collect L-iformetion on J informstion. Such variables were RfLIQON ICatholic vers us Otrr/. as,:ec+s of ..xr,rnure. ".`x `irst is t.tie type of social activities vluch IK70rtE. nARPCAL STATI.'S (single versus ot.her)., arA :u,TeIIt OP CIGLRET[iS . ar•t corducive to exQosure to p.ssive smokin5t senmdly, the frequency %w _1 vtueh s::ch activities .ere cazT:ed cpY: and ttirdly, the stages 4iKEWp6Y asnnq the fozmer sokess. In the second step t+.o o variables vere added in otder to carry out a•fuil' .djuatnent. ' ther Riese ir. life (10 year irxetvalsl ir. which such aL-tiviuea werp casried ouc ~:ables wsr±: esposure to any of -he C!fflll6'ALS (never versus ever wt_h _hat oar-,.iclaar !requenc/. The infcrmstLon cortveyed by these espoisaa) listAd in itear 202-.218 in peqe 176 of Appendix A, and -_.- asoecs o` x.osare .et u.-ve,r::o-4 :n a sing!e _-odex _',at 3Q- -cRSU.-*n to '?VL. fC'Sf OR NALIA'PICTI 0!I 4M J®' 1 pa?a 169 ir ,n,.._.........

-~5- Al;pcxi:x A1. In a.fdiaorr to •.hese, other relevant variables were r cocuidered :or inclusion, but a cleser eramuiatlon • of ae' distribution arnnq cases and oortsols shoaed no siaruficant ciffeaYnces. 'Trse varijoles were: Dfl?OSLRy TO GLEm OR EX1flLLJ5'i F1t25 faQ1 BUSF-1. 17dR7C5 lR 1x7UIP!EM : EXP06tRE TD A GAS r-C0lQTL iut~",c AT 1DkC: t-s50SiRE TO MtOM3E N'?.TV6: arrl l76?Obl]iL5 TO A OD1L Qt t'cXI0 4iWc: (per,es 169 srd 171 of 1:ppendix A). we 9elieve tnat our 5est adjusted estinetes .ere tAose ohrained a rhe us-!usion of t.hc first qroup of variables, since tkey are based on the most reliable inforsatian. In the next chapcer, therefore, we only presux lir. add.itior: to thi cucie estimstesl the results of the ao-wl led 'partL l' ad7ustsent. 11 AWUXSIS 4he e'»ioe of statistiol analysis terlniques wss primerily determuned by the deeiyT of the stady. The latter eslled for the use of tedniques capable of aocontiaq for the ertchinq. as well as for tec"su.yues capable of dealing rnth caulcx situtions in Kuch ~ra:i,ces - in add-tion to *he et~osure varlables -- could `x taken s:mil!aneaasly into consideratim. 3efore xplauu.nq aare folly how t•'iese !echiuyves •+srT aoeli--i. a is pertu:m• to aention thet s:apler tec+vtiqur! •mre a'.so _ ec: ._s_ ~ l:;r^ ;..'ti :hareeterstim of the data cvi:ectedi l:niva:iace SCt%SCCzoz ..,._..~w r. stacistics ard/or sinole frequencies were obtained for alnost all of the variables for %tuch we had inforaeuon. Such exploration of the data also included -- vfrm aproQpriate - the use of stem and leaf or bmc-plot diuqrasn to detersune the nature of the distribution of the variables, in both cases and .x:ntzols. 7Sie ccaparabS lity of :Jties•_ tvo qrcxps also rxeded to be established and several staLitical tests .ere perfoased to such an end. In the case of the varaacles that had been used to perf- the aat_- l:r:g (AGE, S1~7f, gaSNG HISiUiY. RFSMES+M AM 7•.FE OE FN!'F3tt7IF11), we )vie.+ of oourse that no dif;arencip-s were to be fo:md. ei t there rs no certainty that eases a.~ c„tmis wuld be equal in other rrspects. Sociodesoqrapnic vari-ables such as nueber of uears cf s<3no1 attertisr:ca, i.-Lcxve, ethruc qrciup. coirrtry of ociqln, nsnber of timrs mrriad, and othe_-s, were m:oared using either peired t-testa (cnrttiniais variables) or chi- sqiare testa (htNm.r's test for dicotlonoues variables or dii-scpare for cateqoriol variables with tmre than tw levels/. In the sane fastuon those variables that vse viewed as potential contourders were oanpwed first before decidinq whether or not to include tnem as confou.ders in the final analysis. when applic.ble, the ocnparison of the potentially oarnfo:ndinq variables was done by trestirq the variables as both categorical (for aanole, yes/no exmosure to gas ccdk-'nol and eont.inroiLs (for emnole, rxa~er of vears of e)pasure to gas caokinv). ltius, the decision to include a variable as cmfocnier was followed by the derision on .hethrz= to include it as a ceteqoriml or a_ a continnus vrriable. lnqistie reqression was used to determine if there was _vtdenee to asume lirearity rn the effect of the ?ocent_al oonfoundrr. It sry be useful to advance :nv that none

-47- of the potentully cvnfaariing continrous variables (with the exception of utixel s`.cwed indication of lirxarity. '1lrsefore they uere included only as categorical variables. 1:.1 LOG,STIC 'nF)r=:QN tQt f91?C® CASE-CO+SiOL 3I[ID1FS. Nolford et. al. (87) have publisded a sethod to ualyze psirrise ntrtad studies based an the linear logistic model desQibed by Cornfield et. a1. (88). As pointed out by the authors, this method sha+s several fe.tures that sake it the best available choice to analya data sets sos3h as the present one. First, it allo+s us to obtain the estSmte of effect for one or scre esFusure variables While controlling fa variables not cauiderad in the m.ttunq. Secnnd, it can provide estimatas of the effect ass,ociated with differ-ent levels of one or sora categorical espoeure variables.. as vell as t_'x estinates of effett for the levels (tr+e unit darqe) of one or nore cm_iruuu exposurn variabies. t4d third, it alloys us easily to obtain estiaates associatad ratA the different levels of the setctunq variables. In this adaotation of the logirtic nUdel the regrhsor variables are rerpresentrd by the ease-ctntrol differences of the value for Jrn exposure (or cmfouriinq) variables, the intercept is not included, ard tne respanse is alwwys positivs Wh.t is. for eacl value of ttr :egresss varirble there is alvays xr c:K. R arc are ~ctt.-rll. 1!. cesultir.q esusstrs are the maY,mm conditicrnal likeli)md es..inates of the ParaeEtels. In the rrexnt analysis I used the pcogram for loqistic regression available in Q.IM (89). 'Btie unalysis of eecfi of the expcsuue varipbles pra'ceded very auch in the sme .ey. 'fhe followuq, therefore, is a gensral deacription of the atepp followed to assess effect of eadh of thrse eapoaum variables an the risk of lLTq cancer. 11.2 L3MCSlRE AS Cl1TFDORIOII. VAilII18[.ES. I11 inforsation on exposus:e .as ool3ected in sudh a..y as to a alld the creation of variables that wuld reflect such exposure in a cor.tinxxu scale. We considered it useful, however, to look first at tne effect of the %xpoeuae variables as ordinal categories. For that aetter all variables were transfornad into nev ategorial variables. The cit-off points for the categories were cMsen by following aorrvention (e.g. p.cks of cigarettes/day) or by :tferrinq to what ot_5er authors have used in their stud.iw !e.g. 3'richepoulous cateqorizaticn of espoaure to cigarettes in a lifetime). Once the categories had been created ard the infornation suasarized in tables, the analysis praceeded in this say: 1) :fie model that assunes no associatian bee%+een risk factor ard dis-ase wes fi::ed to the .Utu. "us is also referred to as the model with no parsneters or the model of total synaetry.

9t= _43- 2) The loqiric model that conside:s espceute as an ordual cateqorical var:able was fitted to the data. lheae aQ'els prvvided estimat.ea of the effect of eaeh of the levels of ejqon¢e usubg ze,m ex¢usu-e as the re0erece cateqory. 21 In ors3er to test for a lfnear trerd in the levels of the cateqocy, a thitd model •v fitted. This incluled 3s the reqs:sser variable the case-ccnurol diYere~cas cf the wid-points of the otsqory intezvsls. 'ihe ertisate obtained represents ttre slope of the fitted lioe. 4) VP order to test fcr tlae siqnificar= of the oue:all affect of the exposvre vsriable, the model abta_i_ne_d irr 2++as ammpared to 2 _he model a5tu1ned in 1. Rt+a rasultinq ditferesice in G tests the null hypotlieses tMt all p.raneters in model 2 are zero, or sytivalently, it tests wt+etAer sodel.2 provides a better fit to t!x data than the model of no aaaocation. 5) In order to test for the significance of the linear treaai, model 3 was ctnpared to model 1. 6) ^* statistical siGnificanne of the parsmter estiautes ootautied in 2 ard 1..s assessed by crnparing secn est=ptR -.u iu standard error lNS1d's tests not slw+nl, or by est:aati-n, c-eir 956 csfidence intervals. -so- 7) violation of the asssstpclon of the aQUistency of the odd9 2 raum - cleclced by lookuiq at the value cf G obtained in model 2/'[e:st of puasy syimietryl. 'Rwe assnption of the eauistency of the odds ratine in made iadrs the adaprAtim of the loqiatic model usad for the analysis of stched stndies. 81 In the one case that t~a linear trud w.s found to be si9ilifiOlit (etp0st¢e to p63aiVe l~ in the lloUsehold), the nect higher onier model w fitted to tAe data. 'ltie quadratic tesm Sas reprexnted by the ase-cx2iol diifezenoes of squares of the eqd-pour~ of the otigozy intesvals. the 9) A last regressor variable .ws entered to corpare the risk asaociated with zero egmsure to the ZsAc aseociated with an,v level of e1 osure (nane versus we espoare). tk5ne of these sodels shoiwed any siqnificanrti ard s.ince they provide only a 'ooarse• assassment of epos.ue, they are not presented in the results. 11.3 CQ=M AS A CZNt'II2AU5 v11RIJKffit. T'x iact step in the analysis %as to treat the e~xposure varyables as a oan-,.inuus. It .as believed t.hat sueh an appeo,adh could detect si7n:fiont effects not unoaw-red by the analvsis of the exDneure as

-s)- categories. At this point it was also ccnsider4i pe^-Lnent to look at esuretrs for speafic sttata of the study population. P-evious studies have so99ested that the effect of passive smd" mey vary bet,aeen the soms, or by type of interview or accordirrl to sKokln9 h:story. 71m setdred design wuld r+ot allow ue to obtain eatinetes for the ®in effects of such variables (by design ;hoee effects have been •et to be zerol. but estimates for each of theiz levels coild provide useful infozsat;an an the relatiornhip between the risk factos and the study disease. Also, previous =:~aarch rork has sugqested that pessive aookirq asy determine the type of histoloqzc tiaoar .deveiooed by the peCient.s. 'flierefore, we rnted to see if the msqnitude or direct.ion of the effect varied acmrding ttistoiaqic ty-e ornside-'1. to the In sumary, the stepe taken at this stage of the analysis were as fol lcws. 1) A logistic modei was fitted for each of the exposure variables. aodels provided estimates of effect for the sc;rca :re in questlon, taking into aorort all 439 pairs availabie for analysis. :! A;oq;stic model was fitted for each of the exoosure +ariables and their interaction with the variable TYPE OP aNf'fF.Vtf+. EM^ thex imdels estimates for effects were obtained for SF1F- PFSiOrOETT-5 ard for N+!>OGATE RESPQOE3+f-5. ~~~~~~CMoz ~ ~... .rr.~..d. -i2- 31 A loQi^tic r.cdel -as fi*tnl for each of the exavsure var:sbles and the, interactton with the v_*:able 9UIC3-[G fffSiORY. EYvo these sodels seoarate estise;.es were obtsirrd for NENl3t 44R7tS and fF#L4A 9•DMRS. 4) A ioqistic 4r:ie1 was fitted for each of the expoaure variaales and t.'rir urteractions with the varlapl! Sm T'iese rocsiels oroyided estimates for FD1AIXS and wUFS. s) A laqutic sndsl was fitted for each of the eiq~nsure variables and ttheir int,e--actions with the varZAble FIIS7C1G=ZC 'f^_'PE. Separate effects were obtaumd for EYIDf3mID ASO swS, CFIl. CAkC1PLMi1 and aDEMp1i~2['!V1 AND Off4RS. 6) Tests of the siqnificanct of the effects of s11 cf the abo" pnraieters were obtained, as ve11 as tests of the siqnifiance of the interactions. The latter are equiwlert to testing the null hypothesis that the effect does not vary acc¢d.inq to the levels of the stratifying variable. 7) more crnplex nrdels includinq the espos;ure variable in question and two, t`.ree and four t-g)-.ay interacttoas were fitted. 81 Ftdels with the expasure variables and all the poesible w+binations of three-ey interact.ions were fitted.

91 Models ubteined in 9 ard 9 V.re a>ugnret to each ochrt as well as to tte sinplest audels abraind :n 1 throuqh S. 10) 'lhe varsbles sel.,,ted as poGential oonfounders were included in the sodels r++rai^ed in 1 tluvuyh S. 111 A sinqle srrdel that contauied the three sutually exclusive e)powae variables lhuusehola. wdplace and social expo4ure) and t.he corsfasders •es [itt.d to the data. 12: A aodel ras fitted to the data that contiined only the tvo variables with sigrifiant effects ard the eonfouaders. ztVzSCC%o% ,J., W%A AA Q91PTF7t ilftFE RESLt.15 This e`mpeer is divided into thrse sactions. Yhe first is devoted to the descYiption of the study saple in tesms of socio- deroqra,phic chsracteristies. In addit.ion, it ehr~i the distribution of the uariables used to satch caxs to oa2zots ard presents the distribution of cases regardinq histologic diaqroses. The seoaxl section examines the distribution of oertain variables not taken into acccaunt by the metdhinq pcoeess - variables rhiclf. by their roase, way be cuuidered as potestial confourders of the association between passive smkinq and lunq cancer (for example,. pipe and cigar ssakinql. Ihe c'tizd and last sec2ion is suodi*=ded into four subsections, each one reporting the assessmerrt of the assoeiation between lunq cancer ard a di!ferent measureeent of espasure to passiv- siokit;.

-»- DESQiL°r10N oF 1tE SIUOY SArPL Ati) DISIRIffiFi'ICN OP'f`.E MkMIiI1'L V>1RIJB:.£5• 'me results presented here +u'e basea on the scatist2cal analys:s performed on data pertainunq to s:9 msrcontrol pa.irs. 'lhble w 1 presents the must relevant deioqraphic cneractesistics of the study sub)ects. ey design ii.e. s,tohunq!, the sex dist..'ibutian ras set to be aoprox.inately even. 'Paisr there are 218 (E9.71) ferole mse- cantrol peirs and 221 (50.31) pairs of srles. In the saaz sanner. aqe, which .es s.t--red within a.- five ye< age range. caused the mean age for the cases (67.05 years) to be very cloee to the mean age for cmtrols (68.13 years). As is ..11 kna+n,, rost lung cancer ,a.cients aaquire ths disease later in lifa. This is likely to be a reflec-_ion either of a lonq latency period or of tte requiremmt of a prv/onqed e,,"ure to etioloqic factors necessary for the developme+t of the Zsesse). Morewer. 771 of our case control-psirs recorted t'erselves 60 years or older at the tise of disqnosis. By the ume the interview .as caducted, an additional 3.0 1 were reported to fa11 into ttlat category. 'ire latter inQea.te %o&s ob+iously due to the ume elapsed between the dlscharqe diagnosis. tne obtaining of cansent `ron the atteridiuq ohysician, the .ontacting of the pa,ierK tum- or he:sr:: and. `:na11y, the ar:anqe+r•ent and tvrdictinq of the in'ernr_ey. Srr.or 3tSv1v21 >Mr+g lunq canoer cases is k.•nw. to Se poor, and sir.cr ootav.:.-q t-`?e infonrotio+ directly fra the patient tefore death ws eonsi.:e:ed cr.icial for the quality of tne data sax,9'rt, we sied ro ua_o the lapse oatre±n C.taar,oais rd inter+:ev as shor_ as ooss:nle. As a crxsr,s: cc. the eean nurcxr cf days t.-.3z r.ransoi:Ei7 b.:~.rer, _~ MzeCCzoz -56- tw evrnts .as 147 :!ays. 1TUs a! 1a+ed us to cviduct direct incerne.+s antn 296 of :.he ,oa•sents /67.0i of the total). As explairad :n the prec'diir) section. r.asrs and ca,trols were also metctied an type of ir(ternr+' so t.'at the saae prnpoetion of direct ir"-rvievs applies ro the oontrol qrup. Orne hs+dred and runery seves pairs (a51) aere omp-ised by neeer ssokezs and 242 (55%) by former anoker3. 11s .ould be a¢seeted, eost rruer ssek.ing pai.rs %are fea.le ( 75. 61) arid .xt smkiny pairs .,ere srle ( 71.51) . i!r ssonkirsf histncy-ser disttibution .ril l be important rtwen iMerpcatinq tAs results of the espow¢e variable.. It nsst be C°"embered•, therefore, that w>en p='eswtln4 s.sults for f®les we will be dsaliaq snstly with non-emokers, ard when presecttiny results for msles we will Ge referzinq to a subqrajp oasposed eisinly of former lffrJkers. cQ+versely, eatinetes for never sndcers are Ensed an more females thsn nales. .rd estisetes for former srtokers an oore s.1ea. In evaiuatinq iirteractions we sust also keep in mind that the estimates of effect for specifie subgraps will be bsed an a varying nsrber of pairs, and therefore will have different statistiol power, dependinq on the particular intersct'_ons being etiluated. thus, for inst..nce, rh estinates for rreer smycinq sales, will he caleulated using smaliet nurtm r of subjects trun the estimate 'o: fonrer scok.nq nsles. Over 901 of both cases ard omtrels were vhites. Bladcs csnsrltuted 3.61 of the cases and 2.71 of the cmtrols. Pb in:ors.tion an e sric ora.p was available in atvst !11 of cases and

-57- crn:_oIs. M>t patinnts r-pored cheneelves ar hav-_-vq Ixes born in the ;duted States (881). In addittion to this .xxurtsv, 22 ocher cuntri.es, aostly r7urooean, we.re mentioned as place of bsth. Hane ot tlre d.lfferrroes in the distri5ution of ethnic gctup anational origin were statistical:y significart. Fc%er cases than cmtzols re4orta3 ever to haviiq tx-m nar.ied :87t +ersus 92%). Similarly. at the time of t1r_ intesviev, fewer cases (6091 were cepor_ed to be nerrlad tnen aorr_ols. Wiile these differenoes .rse statlst::cally 4iyni!icoint (Qii--square - . 11.46, 1 df) the nurber of timea being nsrried or tfre osen cuiber of ya+ars Curation of nnrriaqe vere not. Significant differertixs of religion %ere offered irf the presence of toce Catholio (5411 asnnq tts cases than ammq the tcntrols (42.0 1) . and by oore Ptuest.nts anmg the tvntrol s( 381) than anonq the cases (321) (Chi-square - 13.67, 4 df). Inooee `w.s fosd to be significarny higher for the cases than for the ur.trols. 7his fLr4inq suggests that mtc?unq on coanty of residenoe does not seae effeRive to centrol for eoaruic statue. Ho.ever, another variable %hicn is generally eauidered to be an indicat--r of wcior--wanic status, i.e. xm6er of years of sdrool atter+dante, did not saov any significant differences for cases and oontro:s. :..e maan mm~r o: y~rs of sdroo: ac_enr~.ance for -ases we. 11.44 xars and. E~r contso,s. ?2.7 years. 'I`w: dist.ributian of hiscoloqic diaqrtues resultutt', !:cm the stuly --viev indicates trst t'ie rta;orlty of cases in our stu3y --ze adenocari-ams (2i2 50.6l): !oides:-aicd earrirana cerst:t%:.-d U-VzSXCzoz -59- TA~ No. 1 RIIh/ANf CfL1FtACf~.RiSTICS OF THE ST'..DY POFLZATIOH FJi56 df?7DICYS FE7*J1IF5 49.71 49.7\ MAIFS 50.31 50.3 1i[74 NE71N 1CE 67.35 68.13 IN 1fpVt5 SM)LTTG HISIU4Y• fffVf3t 9IMIt5 5.01 5.0% FUR6FR 3DIO:RS 55.01 55.01 ^!Ps OP IMIfRVIEN ~ SII-1Pf3F[_X06YI'S 67.0s 67.0% SOF1KX;AIE RISRIVEM5 33.01 313.3% tTfPRC GRCi1P NR+-YtiTl'E 2.7% 3.6% MffTE 86.61 85.61 mU0UTI0d !fi'JV+ lR.Tl1lF OF ymit5 11.4 12.7 OF 9C{JDL A:'If3DAKM PLIiLE OF BDaN l1NI'17D S1711ES 09.71 86.81 O'[3ffR 10.31 13.2% Rf3.IG IAN CA2IDLIC 54.01 42.4% OI7ffR OR IQ.E 46.0s 57.6• HISIOLOGIC TYPE FSIDt7erOID 50. 61 ADENOC,IaC.?4IA 25.1s LN4:;F CIIA, l0. 39 S) W L CTZL 7.-1 T.px3t 6.414 ~ Mac-z~nq vaziables

-19- -60- about M+e auarter cf the patients (110 cases. 25.111. larg~- ard snall were metdni with cuitrols W» we:e former smokers. ITUs Fcocedure, cell accriunted for 45 110.31) and 34 l1.711 cases. Finally, 27 hcrevet, did not necssacily ensure that the psr•scilaritles of the patients were classified as 'ntbar' histologic types. k* want to smckuy habits of the forner snnkers - such as irter3ity arri emchasize that this cateqorization of cases acaordinq to histalogic duration of cigarette smokinq -- were comparable aovag cases and diaqrnees is based on tne diAgrnats pcwided by our study review, controls. Nor did this procedure ensure that exposi.-e to other forms (exoept for the five cases in which the petholor,y slides oould not be of tobaoro was siaular in both qruups. Ztsis. the e)q:oration of the otrtayned). Tus di.stribution is very siaular to the one derived fron d3triLwt=_m of these vari.ables seeae in order. the initial hospital dia4nuees- The overall aq-eemest between the latter and ou study review was 94.01. with a Kwppa value of 0.864. 2 DISIR:.'PLTfION OY P4rDlf1AII.'f CQ~RG VARL181.fS. Scse of the soc{o-demoqcaphie variables al_eady deseibed fall itr-o the wtegory of potentially cuffoundin9 variables. Nanely, earital sutus, income and reliqious affiliation. The ss9nitude and nature of the differences betveen cases and controls teqardinq these attributes have been pointed out. and the usys in which they have been takern into account in the analysis has t+een presented in the previass :na;.ter. Other potentially eanfoundinq variables such as dssnical. omsational, and residential ex{+osu.^es have been discussed in sectior. 10 0! the sane c3snpter. :hetsfore, this section 5i11 an>tntrete on t.'Y ns-mt&.ing variables not ,vet dlscussed. O` particular uroortant- are :1mse variables related to sadunq habits. As •xQlalreC L-I the previous sectirn, the mstching on srokirq histcry +,>, dorr_ on a dicotharous 5asis. 1!at i3. cases :ho were neNer sncicezs w-rx r.nt_hed :o cor.tro:s ~ro o neve.r saoked. and asc: vho sepo-tea as !uvinS ro6-~d The averaqe nunbez of years the ax-srokinq cases repnrted 'Iaviny ssoked was 29.9 years. This figure was not siqnificantly diffemft fran the ar_.an 28.3 years duration of the smoking habit reported by controls (paired t-test a 1.64 , p-.1024). Norever. the sean rw,ber of cigarettes per day gpked by the ex-adc.inq cases (38.9) was significantly higher than th e mean daily nuiber of cigarettes snoked by the controls (23.82) Ipaised t-test-3.74, pP .0002). 2ttis finlinq was reaoqnized to be isportant for tw reasons. First, it suqqested that if an incressed risk for lung cancer was found in the studv. it might be due to the residual effect of past mdcinq hsbits, and cr.x -- or not only -- due to esposure to passive snokinq. Secordly. it made clear the aeed to control for this variable in the final analy,sis as a vay to remove the potentially c,onfourding effect, as well as to st esa the -eed to obtain sefarate estimtes of the effect cf passive srokinq for aevs and forr.+er sadcers. bmisure to other forae of tomaccn did not s1w+ any signuficant 3itfererxe 'aer~eer: cases ard controls. Both qrou;s had about the sane oropor'~icn of intividuals sho repor*.ed e~.r s.~acinq ciWs (78 cases

dold -61- 79 crntrols). Tt, ime_?sity of this habit w also similar for casp and crntsols (s.6 cigars/day for cases and 2-4 cigars /'Ly for rsstrols), as yas the overall lenqrh of uae 120.5 years for cases and 19.26 years for the ootusols). (~afft and contz°ls wre also si><ilar in pecpoetion, intsnity, and duration of pipe smoKin4. Elgthy t'r=e usn and 9o controls were ever pipe sesokers. The cases snok.d and r.eraqe of 4.4 pipes a day Curing ar, avessq" of 20.41 yea..-s. 'Rhe conL-ols s.os.d 4.6 pipss a day duritq an average of 19.9 ye+rs. 7" appn.FiLte srat'_stiol t.at did not show signi:icant differences for any of thase variables. Only 3 can.s and 1 cantzol rep.rtad having smdsxf sarijwna. This diffennce is rot si4nificant. and ir: any case the ccsparison is l+sssd oo such ssall nuabrrs a. to have oo arsniiytul int.erpreutian o[ to raise aQ+oesn regarding any effect on the ^esults. 3 rtS5F59E.'"n' c! ZO fFFt?:SS OF vASSIVL salQM a+ :u+c amx asrs. In the folla+ini section the results of ass.ssinq tl+e isP.et of passive ssddnq an lunq eaur.nr risk vi11 be presented ac-3ing to fou_ d.f'erer.t ,^.ources of erp-re. T+eie saaces are: 11 srokiiq haoits of the soouse(sl. 2)sTnldnq by all seiwysrs of the loLmehcid !inclusive of spa»s). 3) ss+dcing in the wrlplace, and 4) sFICkin9 that occurs in sociai ciit%vstanoes. sfVz&CCzo% I -62- ).1 91OKX'i NJ1B115 or 741E S:Ollg. !bt epideaioloqic stvdir have do.en one sieasurenent of spouse ssnddJxf as an indicator of oqxoase to pasaive smddnq. In the pR,rxnt atdy thn atf.ct of spouse ssflldnq as saa.sarad by three variables has teen evaluated. 'ihe fitst variable c*++•;dn-the <taily smakLy habits of the spaase(s). 'Pl+e s.cod mnoerns the nrOar of years the spaise(s) sustain.d that psrtiailir ssdnng pstt.rn. Ard the third variable - in fact a aombirrtian of the first tw - reflects the total rssbaz of ciganrttas ssak.d by the spawsels) .hile sarri.d ard living together. 3.1.1 ASgS.9HladP c! itR [MR.T at liTelR aR CIOIRa'rf5/UqY SOM) e1f 'lFiE SPC7JSZ. Table 1b 2 presenrs the estts.tsn of the effecC of the spass's daily sseking Aahits .Rrn ansly:ed as an ordinal otegoriol variable. In ord.r to presa, t cwlts .asily cvpared alth those of other studies, es7' ateqory r:fl.cts the standard raat3er of cigarett.es oorrespa+dinq to a dilf.rent nu*ez of pidcs of cigarettas/day. A logistic r.qx esaion afldel ..s fitted to the data using the difi.rvroes in esposure b.oe.n cases acd sruols as regressor variabl.es (See Ststistiui analysis in the Methads Qaptet). Thr-Axqh this tedtruqu.. oorparisoru have b.en s.de between the risk associatarl with expcsure to each of the categories of spouvu daily sndc.inq ard the referent mtagory renresent.d by zero .!cpoare. :yble

-a- tp 2 pre~nts the estLrtss. staedard ezrors. odds ratios .nd oonfidenoe linits for eda of the.e ca.pci.w.. '»ti highest risks ot..rv.d oorrespa ~d to odds r.tiae of 2.96 (for exposure to spouo. sdcinq of sQe ttrn 10 dq/d.y) u+d 1.23 (for eoqasae aarr.epa+din4 to 41 to 60 ciqar.tts/day). 'its oQrssPonKUM oonfidenae liaits. ta.v.r, ind:cste t.`at thw ealues are not signifieer+tly diffessnt iro tn. enll .ala.. :2noP.cte3ly, Urw of the eaua.t.; in the table suqgest a dea.w in riah for lasq cu+osr for tho.e esFos.d to 1-20 cigarettas/dry (CP-.7f97). to 21-40 cigarvtts/day 101.-.9067). aed to 60 to 90 eiqaiet-.as/day (010-.6196). i.llwise. thw eatlaot+as are not •*a++stitally diffea.nt fias the uall wlua. as stw+ in the swr table. A plot of the fiw log odds aorrarpoodtn9 to e.ch categocy is sho.n in figuue tb 1(astia,us r.present.d by diaeQds at ssidpaint of et.qocy valu.s). rbut of the pai~ts sslgst a po.sible linear relatiersship. In ossSer to taat fx liirarity in the categories. the case-aonirol differenor of the aid-point@ of e.eh at.qory w.re er+tered as the reg:+ssar .ariables in the logistic .ulel. The e.ti..te obcaiird for the slope of the fitted line (aatin.ce-.001012. Wald te.t~.2271 ) rs not statistially dif:arern fras sero. ther+fore conc1.d!nq t.tiit t.here ls no en drrs to suppart a li-»sr respan.e for the wtrgories. 2 rl,r~ carpared to a eodal of total syaaary (G ~ 1.0.29, 11 d`) - tdat ia. to a nrodel with no parametars. that assu.s no di`fe-rences in exposure for caaes ard m+tnls - neither C•+e .cdel for exVosure 22 cateqc;:es (G ~ S.JCa, 6 dfi. nor the .edel for 13rwar trend (G ~ -64- 2 10.21, 10 df ), see.ed to provLde a better fit to the data ( e G~ 2 1.99. 5 df, and AG ~.05, 1 df, resp.ctively). alonq the sase lines, a logistic adel that coip.res xero exposure to any exposure (-> I cig/day) aho~+d naither any significatt affet-t nor any apprvpiate fit 2 to tSw data. .tuan ooopared to the mo3s1 with no pua.etars ( Ls G~ 2. 72. "L df / . '!he analysis of the effect of nater of cigarettes/ day sucicad by the spo.+.e also included tra.ting the exposure as a curtlrana. fbr that pupo.e a li+rar logistic srodel sas fitt.d to the dara. 'D+e odds ratio as.od.tad vith' the spouse ..rkirq 1 Qsdc of ciqarsctrs/ day aas 1.011 ('lable Ib. 3. Its 9% oaffid.raa liadts IQ.966 - 1.151 iirludes the null value ud tl+.refore peovids no e.id.roe of a signifiant aasflclation with the stody disesse. it .ay be pastinent to aentian at this point that wfrn logistic reqreasion is appli.d to the analysis of cQ+tinnun data, the estla.te of effect obtain.d for : particular unit of exposure oam be w.d to nlcvlate the eatiste associatad with other levels of esposure. !br tnstarxw, by aaltyplinq tr,e estin.te for I padc of cigar.cree/day (0.01099e) by 2, 3 or .. ++e wuld obtain the estia.ts of effact assodated with the spou.e suduxg the respective eunSher of packs/day. In the saae fashion., aa:ltlplyirg the sturtid&--d error of 1 padc/day by the quantities cited above vr11 }-ield the variation aasodated vith the .s*ixata at tMt level of e>pcaure. :t should be evident by the way in .fildh these ws*in.tes are nttainod that higher oponuss will altieys be aasoci.ted with :arger estimst.s (u .wll as with g•reater variation) so that the stst_+sticl 'sig~::icance of _he estin.tes dws not ury according tu

level of expoaue.. In the present ese.. the estisatas Cor 2. 3. and 4 Fcdcs dqa.~ettes/day .ou1d all be not significantly difterent from the null values thair oonfider1w intervals wuld include arn a fiqua:e.. liqure No 1 sAors the lur that rafresents Uw relationship Mtween oq~o4ure (cigaratt-ma/day in the X axis) and risk of disease (loq od3s in tho Y azisl, accordin9 to the logistic ardel fitted to the data fca+tinuxs linei. Ayain. the alavst e++tir.ly flat 1L+e suggests a ladc of dow rspcaw r.latimatuF. 'Rr tw other lir+es reprew. t two aahqrops of a» study population. 'fie large dash liie aorrespQ+ds to self nspQ+derts, and the an.ll dash litr to mazvpta respoirients. Bcxh of these lines vaild seo to uqgut elfcu dut to esFon:ce that ara not only ,reatar than the effect for the qrop as a Mnle, but also that w7we in opposite directiana frov arr anctter. liowever, neither is siqnifiontly di:ferent fros the line that repcas.rtcs all subyects, nor are they siqrifiontly different fr+aa arr another. Likevise, the estiatas that dssQibe tAw lin.s are significantly differe+t fscs the null value. 1. not 1tr lorer part of 'irible No ]( lines 4-9) aad Fiqo_ws :.'0 2. 3, and oresent !cr the stcdy papnlation estSps+tes and fitted li.-+es, resFreslvely, ap3orcinq to 91X.Ri: HLSTOFOf. STX ar+d H1S-tU[a'+:C p:MaEls. ihe siynifionoe tru and oQSfid.noe limi:s ir.n the table stsa+ no siqn:ficuK effet of spou2* daay saokinq wwf wr.mt.ed separately for nev.r do.rs and foner s.okars, a..les and l-les, ar,d epidearoid ard adrrocucinah histoloaSc typcs. Moreover, ti+a tests of interaction of the .xpowre varieble vith 9CIQ?G HIS''aY. -66 TMLf Nn 2 C3TIPUIT[3, fpQS RtT193 MQ9s1 Q7fI0CiRQ 11MIT5 FOf( E7M06(JqE TO FIYE GTC80RIES OF dfNRETTE3 VER ylY 3fldCCDlY TI( 3iqAC(3). p(MRFTTES /DAY E3TIPS+ITE 3TMQVID Eaaa t1DUS (aTq 95! la 95! tXi 1- 20 -1361 .1 ti" .7/97 .6819 1.0717 21 - 40 -.09e4 .Z03I ..9063 .6oe6 (.3497 at-60 .2001, .3390 13313 .6336 2.3930 e1 -W -.6"a .e731 .r196 ,0755 Z33Z2 ~• IAS00 1.1590 3.lS7b .2"7 27.706d

-E7- TADIE No 3 0MDJl6TED ESTIMl11{5, ODDS PATIOS, AIO Cg1I0tlQ 11M1T5 f(N A DlfiiAlNiU1L OF ElrOWRf 10 SPC15E SF'101GMD OT 20 CJMltETTE3/CAY. OtOIA ESTRMTE 3UWNtD OD0S 952 ln 951 ULL EAAOR RATIO ALL .01999! .G7920 1.0111 .l655 1.1810 34[CYS SEif .1072 0w 1.1132 .925! 1.33N aESOVOOrts vix*"Ti -.2402 .1543 .7865 .5812 1.0642 . 1[SPOODIfS • NM 970cUlS -.0614 .1302 .9~99 .7369 1.2262 roaTA ~ .0556 .100 1.0671 .6622 1.2962 fEfulFi -AY30 .10" .i>'1l . .7W 1 1 ]070 fIMI3 .OSfM .1200 1DSl4 D l7 I.S]56 W"v'04 .11S7 .137rn 1.1227 .SS70 1.4707 3mkt ML AM40CA. -.0420 .D4e0 .96e9 .T913 1.1619 . vncuc Gtvzscczoz

0 p p r Yy ~.-- Y•---#-•--~-•--;----~----~-••-~----~----T . 0 9 ortzs~Xzo% ° e 0 f f ~ + ~ 1 1'1 I 1 t e -69- .

=71- r. a . 1 1 . 0 ~ J. 11 M ~ .3 A d d d d f f f ~ _72_ TAOLE /ld 4 2023382451 >>WJED ESTIIVITLS, WCS MT106. NO (DWIDENCi LInIT3 fpi A DVIfRfNTUL Cr EM=M TO 390uSE SrfaCIK! a 2'J aWETTF3/DAY. CIROUP ISTIMi1TE 3TANDANO WOS 961 lll 95R tlLL EliRlyl iUTq ALL -.0152 OE31 9" .l369 1:1591 ll~~T7 3ELI :Od3A .0987 1.0674 .a96= 1.3196 CKSPD"%TS aRa"TT •.2681 .15eA :7M6 .5601 1.0437 ~~ooEwTS OEM 9+octAs -A6N .1271 .9339 .7279 1:1961 few" ~ .A217 .1101 1.02501 .e261 1271e iElfAltS -.OOV7 .1107 .9QI . :73R3 1.1703 n.u9 .004 .1274 I.a« el]6 13.0>' 010""Jo -.Oa13 .1637 .9219 .N31 1.3215 swl CELL A. -.0491 .1181 9F.i1 w+~ i awi `~~~ , ,

St7t, AND IR3in[CGIC DIAaJF6IS are not siqnificant. indicatirq that the - estuates for the 'ao scb4saps in +ach of these str_tifyirp variables ara not significantly different tras onw another. Table Wu 4 presents .st.iswtes for A1,L 5UB1EZTS and for aedi level of the va-iables TYPE OF fNifRVIFN. 5MORMC HISTCM• SU( and lQSIOLOGIC DIf(aMIS obriinsd with the logistic :eqresJaor sadel that includes oaifouxiinq vsziables. A ocap.rison of these va1Les with those presientes.' in 'hble No 3 sho+s no seaninq'ful differerwes. -'4W tests of siqni.`icanos suoport the pi'aviaus eonelusi.on of no significant association hstuw+ expoas:e aed study disease. '>2ecm is also no evidenr~s of diffe:sYS in the lirrs reore.entinq the ditfarent subqraW for shiek the efle,t of esposuca is being evaluatad. ).1.2 yUMaF3t OP `;FMS CE ShDUSE 90QNC. t9r assasaunt ot years of e3pvrure to spane srokin4 as an ordinal categorical variable dit not shor evid.nos of a significant as.ociation with the stdy dlssase 1'lable Nu 5) All of the sirx estirates are acoamsnied by test results that are rxx statistically signi:icarr., as r.ll as by oonfidenos Limits that unciuda tns nul2 value. A plot of the estisrt.s in Fi%a* !b. S(diawords in .idrt+oirt of tatagoriasl did not suqq.st a stron9 doM-rdp°vK zelationship either. 'ltw abssnae of ws:h relntimst.ip vs canfiineol ty a nai- •iqfuficant test for the reqr+ssor tariable repceer:'-1.'x3 a lirsdsr tserd for the cata.}mies /estuYa+ -.00193. tild• -.6552!. d= M..3 IsE.r of ya.rs of spo.w sadcinq .as aeee.aed as a aa+tii.+ous variable (11b1e rb.6) it y.s foud that 10 years of spouse SnAin9 VMS asaociatad With an oais ratio of 0.99 (9S\ CLs 0.912 - 1.074). F]arosae to as n~ as SO years of spouue smddnq, roughly the highest axpomire ob.etVed in tAese d.ta. wuid ders.ase the odds ratio tm 0.95 ( 951 CLi .6200 - 1.460011 but again, tf.is is not statistically different than the risk for the nm-exposed. Also ir. 'lable W 6 and in Figures !b. 6-!, the estisates ard fitted lines for the differuit levels of the setdunQ variabl.s and the t.n histologic types are prtiented. As ssan. the poi:nt eatLsstas for seA of these subqroaps fltxtustes het+wen ;ositive and neqatiue v.lues, `+tiCh in no case present evidsrrcs of beirri significantly different froa the tsill value. Likewise, the estias:es are not significantly different Cets+aen levels of the variables g7(, %C7@dC HISTOrty ad HISTJLOGIC DIXNOSES. as .valuated by the inceraction with tfe expoeure variable. Borsv.r, the lntaactl,rn b.e,.en the lattar and R1'PL a D+IfRVIDr w[oud to be siqnifiwnt (la<ld tesb~ 2.17 > 1.96, p<.05). >aqqestinq that the estimate far aasvpte z.spmdsnts is differert fran that fot salf-sespQm6snts. Nsasrttrslesl, as s.rrt.iorrd alrrn. neither can be consider.d to be significant. 'fie inclusion of omfoundlnq varaables in thr logistic t+adel qenersud the estimstes presentcd in TSble !b 7. M inspection of thes. [iqures, for all subjects and by levels of the stratif}•inq variables. .as7pottJ the miclusion eorrvfyed by the ksrd justed estla.t.m that th.re is no aseociation with the study diwse. A1t-'txql this basic conclusion is not ehsnqed, i_ sMuld be notni that

7I< TAeLE Ma 5 E3TIMATES.OD00 NA7106IVO 952 t~If9CE'll IIMIT3 fOP E7WOSUPE TO SIX GATEODQIES OF Il!'REP Of Vf1A5 3MOICfDlYTK 3VOU9E(S). MUn6EC ~ r~ I ~TIPLUE '- ~TIP LUE S1NI>ARD EPpflq 0~05 RATIO 95s l1 95i lA I - 10 I -.0970 .2252 .9076 .6a.37 1.4112 11-20 -..3679 .3395 .6785 .4060 1.1263 21-SD -.2036 .2167 .6158 .5335 1.2475 31+40 .2061 .216a 1.2776 .{042 1.8740 .+ - 30 -.2926 .2687 .7463 .4403 1.2637 . 1 -s052 ase7 WN .leso 2.190 -76- TMLE 1h6 1)1WPlSTED E9TIIy11ES, OpOS MT105. A10 CQiIDElC>: LIMITS f0A A DIffET'iEMIN. Of E7IPGSIJQE TO 3PQlSE Jr10KI lq OF /0 TEMt3 9POUO ESTIMATE SIAMONtO OD0S 959 lCi 95t UCl EPROP MTIO ALL -.0I00 .0417 .9900 .9123 1.0744 911D,ECTS SEIf .0526 .0510 1.0540 .9537 1.164a ilESpvOO~t3 _ . -. I 423 .0754 .067N .7463 1.0054 PISQVi1EX63 NWIR 94OKM -.0336 .0633 .9670 .6541 1.0948 rom s,av+s .00a2 .0560 1.00e2 .9034 1.1252 RlUlq 0m A610 1 A070 BOV2 1.12n /1ALES -.~207 .0'320 .0TM Dbe 1.00+A ~~ .04017 - .0653 1.0410 .9159 1.1832 >PIAU clu ADEr0U1. -.0450 .0545 .9560 .C592 1.0638 16 onKM

ra f . I . • 2023382454 , . a I a ^ Is

~ p p ~ .A-..W...~ •a .. I .a sst7S[T%oz : , . e C ra ~ -----~---------s-------- ~ . . . . . -6l-

-a:- TAe1E 1b 7 Ap1WT{p ESTIMTES.ODOJ MT18S. AID C01il0ENGf LIMITS f 0A A 41if EilEM11Nl OF EJPO6URE TO sra&sIMowe a 10 nA+s - OqOIA [STRIATE STMOWD ODDS -- 951 l,a 951 UQ ENApI /IAT10 ALL -.010b .04/1 .1lM .9075 1.07!! s-mrn SELI .0481 .0536 1.0"3 .9442 1.1661 RSOVaBR= SJF#IXAn -.1325 .0769 .6759 .7504 1.0226 ~Ewa~oextS WAR SfqCiAi -.0330 .0657 .967',1 .8607 1.1004 r0w" >r1~ .0075 .0590 1.0075 .a976 1.1310 11nKtS -.6AA .e637 l161 17111/ 1.1266 f SALES =.C 1 SO .0000 .111114 ,Q736 1.1 OE9 040"WEID! -.1131 .0951 .l913 .?397 1.0739 StWi Cnl j AMICC~. -.0617 .0604 .9s02 .E351 1,0584 6 OT/00 L 'l ssVzSCmoz -e2- r1r irrlus3.on of ornfourduq vassadles cawas a chsrqe in the sign of th..st;n.cew for fsales ard for epLdermoil turor type. Miis ctanye. tiowever, is r+ot acoQrpsnied by a change in the test of sigtiticarxar. 3.1.3 TVITi. 11mE7t Or CIGAREIIIS 9mo1033 9Y 1RE: saoUSE aRING MRIWIID I.T.FE. Althxigh alre.dy describsd in the previous ehspur, it is i.portsnt onee .qain to aiphasise at this point the m.mer in .hich the total rsam.r of cigarettes sadasd by the subject spaselsl during wrried life r.s calwlated3 the tra6.r of cigarettes s.ulwd par 'day w.ultiplid by !65 to obtain tM total nsbar of ciq.,Mts wolaed in a year. 'lhis asuuK ..s then .ultiPli.d by the rsmber of yasrs of marnaqe or the rra6er of years of cah.bitation. If therm w arss than one sprxise. the figures for .ach wsse added up into a single total msasureKnt of expoaue. Ztu rs.ultinq figures can be rnsrb.rs en the order of the hsdrnds of thausands. ihis e.y siaprise the reader of this r.pctt, ha+ea.r, it .uat be .rntions3 that thesw seeniiqly high exsfiers do not in fact repr.s.nt vnisuwl aqewass. Fbr instance, a sub jac: aw._-ried during 35 years to a saoksr of 1 pack of cigarettes per day .ould be e3qo..d to ovar 250,000 cigarettes ove.: rhe course of his/her n+rriad life (20 ciqararttes/day X 365 days x 35 y.!ars 1. Tfle total nurber ot cigaretto ssok.d by the spase ..s first evaluated as a ot.qoc:esl variable. lty ©taqoriss us.d were ofiosan to be ids•+tiol !o -.1we •sat 'vf t9'ichcpvulo. Ln his study of CYeek

-•7)- wnen. 1lw estimpces of the effects tot tne S categories essessed are presented in Table 1b 8. 1M first four e.tismtes, aoczrspmding to aspa.ures of cp to 400.000 ciqarettes, shov a rAmystive siqn, suggesting that the effects sry be cmKrary to the hypothesizad. vie.ed statistically, lw+r~er, theee valtas ar* eot significantly different fras the etill .alue. u dsmonstrat.d by the oonfidence intervals alac pcesented in '!able Tb 8. Zbe fifth es'..vmLe, cx--rsGandiM to an expoaae of snre then 100.000 ^iqarectes (OR - 2.0a6), suggests  arll risk above tAe risk for the nrn-expoeed. eut +vaia, t.vs fiqurs dots nct resch .tatista.al aignificayoe (95f CLs .6532 - 1.6762). 'Rre plot that prMrits the estia.tes of effect for the five categories in a loqarithwic sale (riqure No 9 ) suggests that a test of linear trad should be parforu.d. ttr paramster representing this line.r tred .u foud not to be significant. . a~odel CotrFxrinq the logistic ssodsl fa< exposuae as categories ard the for categories as a lin.ar variable, with the null model of no aarameters (i.e. total s)seetry .rsSell, pevvid.d further eviderne of no associ.tian hetveen espo.ure mid «udy disease. Neither one shwd 2 itsetf to fit the dat.a better th.n the null s+odsl (s G • 1.95, 5 df, 2 ard eG - .0061, 1 df, respectively). Dposure, .ssesed as a mntinuous variahl thmrTh a logistic raTression rodsL. yieldeC an estinte of effect equivalent to an x+i~ :acio of 1.0175 for espesure to 200,000 cigarettes during msrried life (:tble !ao.9). '[his estLete w fouid net to be siqniticantr and, as explaurod in a previow s.ction, irdicatee that arry astimate for hzqher or loer exposuces hase7 on the !asti laqistic mndel will be 4sV%Snt1C%az -84- • TABLE Nn 8 ESTIMAT[S. 0D03 MTI06 MO 9Sx COrr10(1[i ltf1TT3 FOR DFM1RE i0 FIVE GITEOORIE3 cy TOTAI. NUMeEii (F p&R>;TTES JP7oRE0 ST T![ JPCU(3) 0UaI1d ty1i1AIED LIFE. /,uneEa oF pOMFFTES ESTIruTE SuraaD EARON aoos MTIO 95= L1i Vs= ul ~-~00000 -.2Q29 .le9i .J163 .6630 1.162" -200.000 -.13TF .2165 .e71e .567e 1.3372 -300A00 -2130 2130 .235A .lG62 .b091 12e30 '400•000 -.0782 .3037 .9247 .S060 1.6935 ~ ~' W3JS .2441 1.046 .6632 1.6762

~ ~ ~ ~ - ---- ~ . / r r ~ I , / , • ' RDW 10 Y I 6ZLZI £Z09' 9K6' 9601' 4S00 ' I Vxroon im tM6 fOrCl 90to' 1f40'l lZti' 9f90' kpaopm 990111 cvlt .ocrl .eor rar s~>h+ GUII i0[V IL0'1 09G/' LOfA' S3Ah8! I 00rL'1 ritY" L0W`l 1•600' 66f0' >131.re10! I oZrZ'1 Q6LL' M96' 6Y11' L910'- u>Tan ~ a" tl10I0A311 fL60'1 1.L09' 1914' 60S1' QZOZ'- u.~c IZIf'1 Z1•ZO' ZIOI'I 6690' ~960 l13S __ fL'1!f'9fK f0Y 1 1 LLLY' U101 LSLO iL 10 ~r JILYU a(m) mft 1iS6 l3l 9S6 SC>M Q!'uO/YtS 31YIJI1S3 d10tlB rJL13WIC 060'00Z !0 GNIXr. 3SJ10dC 0: 7VIS06A !u 7vI1N3Clli/0 YZID! Cllwll 33/)OIPaJ QNV 'qI1YS C000'CIIYNIaI WLt?YtNNh 6 'a1t ?IYYl -S E!- s.-7-' VC,as(.. l.r C.O Ci -98-

0 0 .... ... . / M! 1~ ~ ~~~ : ^ / • , . t ~ • • • -ta- ssfizscezoz _fire_

. . . t 0 20233 8 0-C!'4 60 TAELE Na 10 AWIAtED E3TIPMTM ODOS RATIOS. A/D EIItf IOEMCi LIMITS ipi A©IiT EAF 1R 1AL OF E7fPCStAE TO 3v0USE 3MaIcIwOt :00.000 ao+RrrT:S. 6ROlA ESTIMATE STANDARD 000: 95! lfl 956 U4 ERROR MTIO '~ .012t :0E0/ 1.0129 .6667 1.105! ~ ~ ~R{r~ J4/R13 3ELT I .0666 .0d94 1.0916 .9170 1,3019 a~svvatxf3 -.1 a97 .1537 :/272 .6097 1.1224 ~~ 99Ot[Rf -.0212 .1226 .f790 .7699 1.2 450 raw" YVK05 .0142 .246E! 1.0390 .6,2e 1.6T93 fUuLFS .0067 .104 1.0067 XIG'2 1]737 nKfs -"16 2190 1.0219 .NS7 136+. m~~l -.1776 .1 T'S9 .6J1r .5931 1.1818 gvlL Ci11 ADOOMA- -.062e .z:66 .9205 :5790 1.4636 I r.onEns

-91• pually ron siqnifiosnt. Also in Table No. 9 eutias<tes for the levels of the stratifyurq variables T7YE OE INI`.3tYIr7a. 9'OlaPIG HIS2dtY.. SDl and HISIOWGIC DLO04CSIS do not show significant effectt , aod in all oisem the rstiasrtes wcse nrx fourd to bs aiqnlticr+tly different within levels of the sase variabla. In-additim to this e.ch of the eodels that irrluded the interaction bet.w..i t':r stratifying variables ard •he exposure variable did not pcovide bettar fit to the data than the aadels for all subjects, or thu: the suiel of total sysmeuy. An lllustration of the lines fittad for all sutjacts, alonq with the levele of the stratifying variables, is sAa.o in tiquras 9 trough 12. Einally, the total nabrs of ciwrettas ®k.Yi &Xrinq sn.:i.d life %ss assassed in the prsence of oasfoudinq variablas. 'ltible No 10 slw the estiartas obtiirrd ard their omfidsncn liwits. in ro ose did the pr.s+nce of oovariatas in the sodal impswe the fit to the data or um+eil any siqnificarK effact of the exposure variable on l:nq arwer risks. 3.2 PASSIVz 9S7QH6 IN 'P!E HUISQ!(Z.D. Cases aM oont.- ols Mere qusstiom.d .bnut the nsber of yaken with wlx= they sn.rd their 12 srast -eaent residences. tcqether with the durstian of aad: cctuabitatian. '!T» rnaber of ssokers in eadi houd:old, sultiplied by the naeqer cf yesn lived in anch particvlar Aa:sthold, and suead ouer the 12 residscws. estimates *h* nurtxr of person/years of sqocuea to passive raki:+q. -92- Inapection of the rav data slm,.d that for srne of the 12 reside+wes - particularly ttose aorrvpondinq to early staqes in life - tlrr informstion on the n.etaer of ssdcers in the hoenetnld ..s sussing. 'Ire data w analyzed in two .ays, each asJcinq a different assaytion about t.tiese sussinq valoes. lydez the first asstnption .e ansured that one smaker hW livad in the Arantlnlds for whid: inforsation ras eisainq if,, that is., t.hera .as evidmo. thatt the sttidy subject had livad in previous or future r.aiduces with at least ons other srdeer. lJnder assaptio: ntzmber tw it w ass:aad that there •as no exposure (i.e., ro ssokers in tl:e AousN:old/ for thosa resideraes in rhicl: infonastio:ry qissfnq. p.sults fzaw analysinq the data under the tw assu:pt.ians are prattially identlol. Hxeves, both sets of results will oe preasnt..d to alior the ra.der to .ake his/her oxf mrrlusions regarding the hsidlinq of missing values. ].2.1. RiSULZS UNDEl ASSUMMON 1. 3tie rewlts of analyzinq han.hold exposure as a ot.qorical variable are shoen in Table Ib. 11. Dcoapt for the estissts aorrasponiinq to 26-5o person/years of exposure (Ot+-.9166, 9S6 cS: .SS10 - 1.5201), all point odds ratios suqqest an increased risk of lunq cancer associatad with haiselqld exposure to passive smkinq. Bow.v.rt vily the hiqh.st exposure - Mnid: also produo*s the Rigpsst astia'.te of .f'aca - seru .vid.ws of statistial saynif:eano. lat for s~osura to •17S pessaMye.rr 2.t5 (9S9 Q.i 1.09 - 7.621). St.ill, the plot of the eir,ht estimstes presa:ttl in Fiqute No 13 siqgssts that a lirwAr sodel sev ye adec,uate to ,iscstle r1e ;.*:atiorLs'rdp oetreen - i cst-qori." of

:.- eiposure and tha risk of diseese. In the usual msrusr, a variable reflectiny limarity Ln the levels of the categories ;the case-cvrtzol differenc+e of the midpoiiKs of the categories) was entered in the regressiai .odel. The resultino .aiel atnroed to be eigtificantly 2 diffnrtra f1cm the wahl with no parasetars (p G- 9.01, 1 df). Sn add.ition, the eati,iete repcesentirrq the slcpe for the fitted lv:e ms fuud to be siqnifiontly Gifferent from zero (estin.tr .O01s85. unldw 2.95 ). ltie cosFarison of the aodel for the ateqories with the aodel for the •lirrer• ote7ories did ent provide evidence of 2 deparaue fros linearity (e Ga 5.87, 7 df). A higher ocder model with a quadratic term w.s fitoed to the data but it w found rot to be signifi,^.ant. :f» aodel fitted to the data to ateeee the effect of exposure as a oontinwus variable yielded estLn.us that sopFrxt the eustenee of an asacaiation Ge!tiesn eaposure and study disease. 37r adds ratio associated with 1S0 pat'sorVpsar's of exponae .ms 1.9%, with lo.er (1.21e) and upper (2.l30) 959 canfidr~ee liidts .ell ahova the null wlue (See Table !b 12). ll,in4 thr same salcl ws can astim.te the odds ratio asmor.iated with a very hiqA #nqoa:a+ of 250 o.rsan/wcs ae varying Eetwrn 1.389 and 5.66, as indioced by the ;wrt estimates cv+fidencz livts. ;Ae effect of :.xuehold swokaq ac=xdinq to TYPE OF r.qR'FRVIES4 dces not zhw di'fereneo-s in *~'~ effect for self-responc;rncs w1sLs su.^ogate reswdents (Ta!~la r.o :21. Hobatier, +rtrn StCPQDC ei25-iL)7Y is -9l- considered, the effect for newr s.nkecs ia not siqrificant. .t,eaTas the effect for ton.r assokers is. This tiniinq is all the eQe pizzlinq G.cause borh estiimtes are almoet identical in value ard because the test of interaction suqqests that they ars not significantly different fzv each other.. 3fw effect for eadh of the tw sexc; is also estissted to be vary close to ar another. In this case. hoble.:r. hoth are siolsifiou¢ ard not differaa frv each ocher. A higher effect of housdnld passive a.dcinq ..s foird associated with epidemnid ard ar.11 oall histologic typ.s (CP-2.l3) t.han for adenocarcirowe ard other o.ll types (P-1.921. lbreoeer, the effe,.-t for epidemuid type is significant 1956 Q.: 1.39-5.761, ah.ree" the effect for adenocarz•inasr is not (959 Q.: .s35 - 2.41). Pi9ures !b 13 - 16 illmtrate the lines for the r.qression .ui.1s for ead: of the subqrtxps des!Qibed a6ove. Tfie medel th.t assezses the effect of hou+.told expos:ae in the prasence of mnfoundinq variables yields rmlts that are different tro>* the results of the modal vitl:out sueh variables (Table tb 13). The estieatm of effect for the Mole qrvt4 ia sliqthly lower /e7t - 1.556). but no change is oC.er+tl in its statistical significance (9Ss eZ - 1.p006 - 2.42). The estis.tis for sarrvqate cesporde+ts, seif- resoordents, forner srdkars, males. teeles, ard epidernoid histologic typs - formerly significant - do have oonfidenoe intervals (in ehe presenoe of eovariates) that include the null value, ard pnint estiastes .Riich suqqest lesser effects th.n the u++adjusted estsnates.. lfie point estimete for aderer.areinana IQR-1.97). althayh still la.er

2023382463 •IIM Al VR9C ' 1tP1/Wm Apn qhlN vd,bUO , 1LOr7 1Sf4' Z416't OOLZ' Si6/.f' v~ rm rnus LSSLS KYC, I 4LZ4Z 4Z9S' S6£0'I 900Ct [O[vl 11091, oL6C oLiY S7riu p9.p,t W101 [196'I Otl[ YUT S)1vNL f01ff ZQtM'1 9494'1 416Z' ZSZ9 La7~n~i ~yp~ LYfip'i f9Y6' 91Y9'1 94If' 59019` Sayaa gyqN 4xfmoe¢lu LIi1'Y pSZO't ISIP t'L KLf ZS9L' uv9orbts uxluoas~/ I`064Z 6Zf0'1 6LZ1'1 SZ91' 69/•S' 1139 5131= Y6Z42 4LitI f994'1 IS1Z'0 9419'0 Tr atvC aom 171 996 Y71 =96 Sm QW011/1S 31vNItS7 d110!!B .01d(3rf1(H )it NI WI)Q4tCiW34//[Stl710'41 !D Td15WE1 !0 Tlli/(31l3ilIO v ff0! StlNll 3RIl0Il10f) ONV SDliw OOm'C]dvuUtl Q)ttfV OMIt1 Z1 '4131Wt i iiow A W1 -S , 1 L0.y0uMwc ~ clinalif+ 9LIlpL ZL601 -Z9MZ LQWp' 09rG 1 .~1 IrL1'r $0(' SOIFI OOM' Y9[Z' LEI-Itl ucrr ocw ol.vl cess, relc oaI -9t1 L91rZ QtiIL' SZZS'I 6tlf' S6LZ' Q1-1m ff9L'Z 06L4' 9Q93'1 ZZ6Z' Liw' 001. u 95ip9'Z Z'f6' SOLS'l :99t 1+1p' - QOZS'1 0154' 1•416' 06SZ' 1QY0'= ( 05-9Z Z1L6,1 6fS!' c'Sfl'1 414Z' 69Z1` Q-1 ( rn=S6 mi 196 atvd soao wan] aavoHVts 3ivuIts3 snA/Nasn3e `QOIlCf10<1 )It NI O/II;b1~4 D f7/11M1ta 1NBIl0t 7tlf'.SOd1(7 fA! ttINII mi0I141 926 OA"MIlvtl L000 'CllvUIIS] II'p1 l19Y1 =96- 1 -Se-

2023382-164 ~S>~8 0 1 2 JOO 0000

W O 1 -99- -100-- 202338246S .` 1 , V \ , a t ... ~S ! O + _' y fY !` \ tf SJy - . ! g •` t zY1~c ! s r2+`' ~ $ !; ~ 90 I '~ ! E6 ,! SWE 8 _~~ ~ ~ so O W a cI a I

-.lcT- -102- tlin the estsarte for epidermoid type /OR•2.321, takes on a borderline saqnificant value. TA4lE 11:L 13 AG.WTfO ESTIPWILS.000.! MTIOS, 3.2.2 aFSIII.15 UtJDER ASSI?47ION 2. Ai0 tXMP:D:ICX LIMIiS FOP! A OIFiEi1fWTUl a EIED(>SU1E ff 150 FEF80N/YEWIS SMOtIMO H/1NE 1qWC?ClD0 9ROtp E3TR7ATE STAWMD OnOS 95X lIl 951 UQ E4AOR RATIO ALL .44122 i253 1.5861 1.0006 2.4200 9r.LCTS 3E1F .3238 .2764 1.3824 . . .0041 Z3766 I[lPOOOfNi .. .. ILMN"n .6726 .3937 1.9593 .9056 ti.~.'•91 ~tavvac>rrs a,cR~s .3753 .3093 1.4554 .7938 2.6685 5157 .3266 1.6746 .8831 3.1764 fENAlt.l .R70 .S2'S6 I S?78 .770 21M nAUS .450 3090 13817 D631 2D986 OWW"0of .6428 .5006 2.3230 .6709 6.1996 SMWL rIi ADU10Ca• .6288 .3188 1.8678 1.0900 3.4886 a Om0a e4b asrlb unNr mu+ielkn I. So 1sri kr tEln1L S9Vz9CCZW! Table No 14 pse.ents the tesults of aculyz.inq Aoanetnld exposure as a ateqar=cal variable .fen AoaueAold exposure w ola+:ated urder assu:pc.ian 2. E]ca~t fs the first tw escSmstes. all suqqest an ineeased risk for lsq sarloer asoiq ttu eaposed. 'Qie plot of the eetisrtes (Fig Db 17 1 also sss to su}y.st that the risk in=s.ses with incx.asinq aposur.. Miile nonm of the estis.t.s is significant (see oonfidenoe liwits in the table), the linesr t:wd is 2 l 0 0• 7.014, 1 df. qoztq fmw the sod.l with no pars.etazs to the nadel with the reqr-essor reprasenE.tr+q a linear tsvd for the categories). 2tus. in mmparisan with ths estinstas abra9ned urder assualxion 1, the present estiaw only differ ie the enm siqnificarns of the estiwte for the hiqhest asposta+e. rf,.n analyzed as a aantiruaus variable ITable !b 15 the estiarte of effect l0it• 1.73 for 150 personlyears of exposure) is slightly low.r •tJvin the ef!ect obtained ulder assurption 1(UR - 1.856). Horever, thei.r crn!idenc.e intervals overlap to a ecrsiderable degree (9S1 Q.: 1.218 - 2.83 .rd 1...37 -:.6r,0. caspec:ively). Sig+ifiont effects were fowd for only t+.e cf the s;:oqroLgs arrlyzed. !ocsrs sedvers (fR ~ 1.770 , 951 R.: 1.0009 - 3.132) ard epidenroid t.istoloqic type (CR • 2.S8S. 95% d.: 301 - 5.137). [nclusion of crnfoursiinq v.riables in the so3a1 (Table No 16 1 resul•s 3n non-auqtiiticant effects for tha.e and the oc.tier suGqcv.ps, as .e:: as for the qrup that oa.Fxi.ee all subjee'-3

-loi- TADLE Nu 14 C3TIFlATfS.00LlS RAT106 AIO] 7SZ CJNFIDENZ LIM1T3 FOi EXPISURE T0(IINi GTEBOIIIESOi 5r1W1N0 IN T/[ 14ME1[L 0D PERSON/YQS ESTIMATE STMl0/t(1D ERROIt 0005 RATIO 95! LEt 951 UCl ' - 25 -.0960 .2669 .9007 .63K 1.5332 26 - 50 -.2604 .2410 .7707 A06 1.2361 sl -75 .25a9 .2.7e 1.3343 .e210 2.16" 75 -1oo .2461 .2639 1.2776 .7525 2.2290 101-1n .1030 .2972 1.1055 .6191 1.%,4d 136+IS0 bNl .160 1./373' JLte ~327/ 151 - 11S 2003 4513 1333e 3090 l•Va7 171. .e521 .4649 2.3446 .9061 6.D6sa •Nwlb uram "Prnotlm 2. Sr tat'tr Awlb ~~~zs~~zoz TAELE Nn. IS U1MDJ133TED E3TIfNTL3. ODM 11Ai1fb, AID CD/fIDENCE LIMITS Fa1 A OIFFERENTIAL OF EZTMRT: OF 150 PER30h/TENtS 3rTdU W IN il[ /WSEhIDLD 0 0R01A ESTIl1ATE STN6,A.zD 0D05 952 LG 95Z UCl ERROp RATIO AU .b492 2146 1.731• 1.1371 26376 >i~iRS 3ELF .5063 .2622 1.6592 .9l24 i7739 RSVVcDRf A»"n .6356 .3754 1./rfA 1 .9045 3."10 as~oaaT3 Kva s+arns .5226 .3192 1.6664 .1021 3.1526 Ecw" lr+acD~ .5712 2910 1.7704 1.0008 3.1316 itTUil.E3 3204 .3102 tleU .9161 3Ai06 flKtS 3754 200 1.7m .9042 3.10m PWMICq ."% .3504 2.58" 1.3006 5.1368 9411 CIIl ADVIOCA. .2b02 .2775 1.3234 .7682 22796 a OflQ13 •Reaulb urdr asaeptlon 2. SmI 1eA /Q det.fh

Y-~ .I • ed I 1 r j a&svzRCCzoz -901-

~~ ~-- f r F F r ! M ~ ~ ~ _- ~ - ~----~----~----'~ . s9tzscuzoz \ \ M M •o,-

-1@9.. TMLf NO. 16 ADJUSTN [STIMATO, 0006 A11T1I2S, AIQ OJFIOE1Cf lIM1TS F0R w 01: FERENIML OF E390SWE OF 150 RfRJ0/VYE*A.S3rtt7CIN01NiME IOUSEiQO • 8R011p ESTC'WTE SU111fINtB 0005 952 iCl 951 IIII ERROa RAT10 KL .3732 .2M 1.4624 .93Q , 2.2560 SS3ZECfS S[U .2518 .Z763 1.3256 .7713 2.2752 WSR7OCNR SAW4aR .5510 .3906 1.7369 .8068 3.730s RtlvoCtnTs WAR .4332 3_ 2 .3258 1.5422 .0146 2.9/97 FVWS 1oreltA snouAs .3186 .3091 1.3752 .7G03 Z620a IDWlS all> >2e 13181 .720 33fl9 Mal[f .497 ]101 11382 a703 2dt12 wcuraq .1696 .1968 2.1590 .OI23 3-7385 fn.Ll ml ADDOX.A• .5691 .3165 1.7667 .9500 3.2e55 a aTwa Mtsau113 uVr aNVepltaf 2. Sn teA far bfMls (OR - 1.45, y5. cL: .934 - 2.258). -110- 3.3 DF061iiE 1V sOKING IN T!W HORlO'LAEE. Nine categories of ssckLg in the wrlplace vere ceated aotording to diifereat leuels of expowrr. 1tr values included aithin ench ntaqoey (cu:apt for the referaK z,ero expaw:_-e) are shw+n in Table No 17 under the h.adinq PQt~N/1rE7W5. 1t:e ilrterpretetion of such usits has been explainsd in tha pctvious ct:apt.r in the saction that ddeiues the espos:ae variables inspecrtion of tha estLestes obeaiird sln+ad effec s ranqiuy fcan odds ratioa as loti as 0.a664 to odds rstiw aboue the nwll value (1.187, 1.186, 1.166). Nans of thns is statistically siqni:iant (see aolums for aanfidsnoa liaits in tAe table). Itw: these estimates are piotttl IFI.q. No 21 ) thsir distribution in the fiqse is scattered. thus suqqestinq ro dose-rsspv:w relationsnip, the aCsenae of whid: w oonfinned by the nan siqnificanoe of the effert for the variable representinq a linear trerd for the cateqorits (estieatr -.000287, Weld tast~ -.2839). Neither the aodel that assesses esposure as an ordinal oteqoticsl variable, noc the sadel that ass.sses the lilrar tFerd for the cateqqries. providas a Dettur 2 fit to the data than the null sodal of ro association (A 5 a 7.83, 8 2 df, ard a G- .084. 1 df, resp.crt.iwly). This further dsniaa evideno. for an essoeiation Oac.er: expomsre to stdcinq in the wriplaw ard lv+q onost. The assessmmt of e:pasure in the ++orlpl.ce as a arntuuous

lA6lE Hu 17 ESTIIIA'E_. OO03 PATIOS ANO !S; tDNFIDENCE LIMITS FOR EJm(15UL TO FIEtiT GITEOONIES CT N10[INO IN Tl[ WORXPIACE PER30N11'AS ESTIPNTE STANpYtO EPa¢I OCOS PATIO 93= LA 95i llfl 1- 25 -.2091 .2006 .6113 5m 12U 19 26 - 50 .1705 2513 1.1859 .7'2f7 1.9407 so -5 .106 .3063 1.1660 .6372 2.1337 76 -100 -.5s31 .4402 .67i2 .2427 1.3630 I 101 •13s .1717 A516 1.1t73 .4669 2d9S/ 176 - 1!0 -34T1 bQ .7067 .2V0! 1.7212 151 - 175 -.M23 3176 A66~ .1643 12870 17S • -.0377 .2141 .9630 .6330 1.4652 TAEIE Nn 16 1RMDJt1,STE0 ESTIMATp, ODOS ItATIOS, NO COFIOEICE IInITS FOI A DIFFIAENTUkL OF E7obSllslE OF 150 REP,x11/TEMti 9r71x11O IN T1E WMDLACt 6ROlP ESTIMATF STAN100AD (1005 95= lfl 959 UCi EPPpI MTIb ALL -.0060 .0104 .9411 .9740 1.0146 SS,UCTs 3EU .0138 -.0170 1.0139 .9E06 1.0463 ats>NPVErts AUMUSAn rasoa~o[Yt3 -.110a .1249 .8951 .7008 1.1133 K~ s~arns -.05329 .06" .9481 .63b 1.0767 Farm ?1o4R3 -.0047 .0105 .9953 .9751 1.0160 FiMAtl7 -.3140 .0017 .1M1 a2J' 7 1.1374 11AU S 001. .0105 1.006 .9E31 I.0265 -.0068 .0124 .9912 .9674 1.0156 SnAtl ml A°O'OCA. .3037 .0223 1.0037 .960a 1.0464 & OTIERS

i W N R 2• ~ H, I i f • • . e W -114- 2023382472 . I • d • f f • I I I . I ! R I

1!~' p r ~----- A. f L f r a 4 al i I :l il . . . • YTMV V\JO% . . . T ~T1~ I I

-117-. iAB:E Mn 19 AOJWTEL E3TIe1sTES, 000S MTIOS, AipOWlDE1[i LIt1ITSF0AADIFFEFIEMFMIOF EXPOSUfif a 150 PEAAdVTENl33+101UNB IN TI[ WOLKYLACL fi'OUP ESTIPVIFE STIWOSiiD 0005 95% LO. 95s UCl I fAROR Mi10 ALL -.0101 .0106 .9900 .9797 1.0107 9"CTS tf lF AESKKAMn .0105 .0170 1.0106 .9775 I .0418 1PROSan -.1447 .134S .e653 .6648 1.1263 acvwx4 rt! Wm 9iocm -.0900 .06716 .9139 .0012 1.0425 TCW" ~tOKs7lS -.006t .0104 .9919 .9719 1.0124 it)MItS 'D6/) .003T .*f/J 1009 1.1245 fMIES r_0091 .0105 .9906 .9704 1.0113 "30 1 0185 -.0_ 15_ 1 .0222 .91H7 . . tTtAV QLl AC(W"-a• -.0t 13 0120 .9886 .9656 1.0121 t OnEIS t4t%SUCzoz var:able sipcorts =+e fiM:nr,s dasribed for t,•.r catcyor:cai exocsure variable. 'Ftbe o3ds ratio asscciated with an expcaure of 150 person/years %as 0.994. vit.h low.r (0.974) atd upper 11.014) 951 cm:idenoe limits r,hst include the nu11 value (Table do. 18). 1txewer: examuation of. th...sti..tas for the d iffera~t subgrouFti definef by the le.ele of ths stzatifyinq variables furt.'srs suppoct the lack of evider+oe for an association (loNr part of Table N4 18 and Piqaras No 31-2a1: lShst+is., assesaAen: of the effact of aposuae o++•A+ned ftas cdsls th.t includs the oonfouniinq variables do not pcvvida sutisticaiiy siynifiont ert.iaatr (Table !b 19). 3.4 t9(POGURE 'PO PASSIVE 30WC IN 80CII1L CDtCLPS517MtM. Eight eataqories correspondinq to diff.rent lsvels of w¢ostse to passiva smakinq in suoiai settings .ere oasFrred to xero ruPoKa+s throuqh the fitting of a loqistic .odsl. 9uch a sodel prwid.d a 2 taetter fit to the data than the sodel with no parae.tars (~ G ~ 26.41, 8 df ). ltr astL.tes for the tw lor..t ®cposutas wres3pord to adas ratios of 1.838 and 1.061 tifbl. No. 20). '!Ae suqq.stion of ina.esd risk !or ltnq eana.r for these ot.qorip of exposure, hor.v.r, is not wippartad by the wide oonf!denoe lisits Chat include the null v.1ue. 3Tu air rsi.ininq stSmtes of eff.et - all non siqnifiont - do taloe n.qativt v.lussi they also sss to indicats a strorq linear tred in that directi0n. fYqurt !b. 25 ahws t'u points suqgestive of surh linear tretd. A test prvvided the ne.oessary stutistiaa: .videnca (estlnate - -.022d3, wld- - 4.380). L11

sUnnary, by attinq a logistic nodel to tti dats that includes asposure as a eateqo.inl variable. .ow fasd eso things: first, that the resulting acdel fits the data besosr thsn a eadel with ro puransltars leven thwyh eadh perasetar in the firat .od.l is not siqnifitantly dtfferent frum zero), and seooni, that ahrn ataqaries are eade to reprrsent inasaainq Jo.es of espoRae. they sha+ a linear trend opposite to that hypcr<.'asiisd, that is, a decrease u+ rislc vsth incraiau+4 axposure. 2`w analysis of the irdex of social exposure to passive ssahing as 4 xrrcinioue variable does not ch.nq" the picture (Table No. 21). F]Qa+ure to 20 units of the index of social wpoauw is ae=rprviied by an odds ratio of 0.699 (951 OrnfideYne linitsz 0.5202 -.7850). 7fie estiartes for the differwit levels of the strstifyinq variables are all close to the value ciu+d abow, with oanfidance limitu equslly excluaiw of the null value. Fidi of tJrsJ are shom in the lowr part of Table IU 21, ard the iSxs relzessnting the linear trerds appear in Flqures 25 through 2!. Furthernare, the inelusion of cvnfoudinq variables in the acdels fitted have very little iapct an the valu.s of the point esr_in.w /"sble Nu. 22). 'Ry similarity of adjusterd ard unadjusced efftimsccs is also reflected in ouetlappuig omfid.ra.•e imarvals. In tw 'Lwtu+ces, horwver. the int.cvdictian of aanfou+diiq variablee had the effect of yie:dinq ron-aiytificant estia.tr. Fhr susr=qate respaidents tha coniidenoe liaits a.-msd the odds ratio (0.957) include the aull value (0.6424 - 1.426). in the sms.e wy as do thr oontl*ree lim'_ts for thn S 4 fizS C C z 0 z -120- 1Ablf Nn 20 [}TIn/1T[3. C003 MTIOS A10 95i COMIOElfI IIMIT3 FOR E_)PORJitE TO EI(9R CJ.TEOCAIES ff Dti4SiVE SMCKII6 IN J0'JAL 3111i1T10KS DIY45URE 11OE1f E3i IPMTE STfItYWD ERAtR ODOS MTIO 952 Ltl 95I- lRl I- le ~7 l6SE 1.6i60 33!<1 9.9! 16 11 - 20 .0617 .S6M 1.0637 .1954 5.7689 21-30 -.1034 .a197 .l01e A706 4.7684 31-.0 -.4946 s63i .609! .114M 3.2606 e1-30 -.5363 .e635 .5899 .1077 3.1TT7 31 -LO -3000 JldS1 .iMS .1112 330Q1 61 - 79 -i aup .9371 2]IIO .oJ67 1 Sr?'! 70 • -9.a>'oo I.4saO .1541 .00e9 2.6~93

TA6LE Nn ? 1 UIMAAIST:3 EST1ri1TE'S, nDOS MTtal. A/D tr 61DENCE l It1IT S i(TG A OIFTLREx'lAl S E)CPOURf rf 20 U111T5 IM T1f IM9f]( EI ©POSl>tt.` TO 3XU\ fASS1YE S'1e3>;I1a, Eii(7p ESiIPVATE 3TMIp1~.'1D ODOS SS% lQ %i Ul E!l/WP. MTq ALL -.4476 .1060 .5390 .5202 .7650 91dlfC 5 StL! acgxacem -.1614 .1240 .6J67 .4493 .8119 strraumtt -.4386 .1945 .6449 .4405 ."At ~ ~ 9~+acfns -.36za .1591 .5696 .4170 .7761 ~Gt1^ot g+Q>AS -.3510 .1420 .i640 ' .5329 .9300 1EINlif -i6P! .1674 ~1t1 3009 .7110 MAU3 -I7id .13e0 :7370 3706 Vo07 04013"Mi 1 -3604 .176b .6974 .9914 .9e90 slvll ML ALOWCA. -.4926 .1300 .6110 .4736 .7884 & onEns -I2z- 2023382476

W h O ~ W¢ ddd~". ~ ~J~ lf . . / .. ~ . ~ l . s i R sR r• Y fv i s s i a .J.. .... a p a I , / f , f . f If • ~^ f f ~ --- 1"f 2023382477 I .

-125- -126- 2023382478 ~ TADLt Nn 22 AUJU3TO [31ItMTi3. t>DOS MTTLS, AIO COIiIDE?CE LIr1ITS rCM A DIrfT%1RU1L (1< F21af>StAtE ff 201M1b IN 11E I/QUf Ol [WWUlli T03[X]AL PA:SIK S!'IMIIb "lP ESTIMATE SIMOAqO ODOS 911 l.IL 953 Ul ERAW RATIO AM -.46'YI .1107 AM .i034 .71H we,ECTs S[tr Ws+ooXxrs -.1016 .13(T6 .i170 .4753 7l71 siwoAn -.0439 .2034 .95i1 .6424 1.4259 1NSVOratKT3 NMA SMCKM -.5310 .1626 AM .4715 .l08A raw" sIoclns -.4141 .1513 .6609' .4914 .8E90 rvuL[S -.i.oa .1M !oQ 'J.~. a0s+ rUllfS -?990 .1456 ,7414 ~61 AEQ3 O~ocpnaDl -.30z0 2418 .7393 .4603 1.1676 s+,.u aai Aoaoea: -:3%3 .1&43 AM .5070 :es3e a 0naas

esc:mete 'or zoicsi-nid 1i•r,nlo3ic t}w Jues ;oA - .a333. ?i11 Grnfidence lini*3: 9.4503 - 1.1881. s4VzRCCz0z O91P1I7t PaA . DL9CLSSIC7r Th. hyF[rtlrsis that lunq cancar ssy be etiologically assc+a+r+ with wqpowr. to .nviras+ental mtsooo ssob has anly rxc.ntly r.o.iv.d •atvr.loe frtsA epidaiclogisCs and otlrr scientists. Ttw possibility of such an asaaciation had b..n raised in the early 1470's, but it w not until this d.eade tAat the first pcpulation studin addraesing this issue .a:* publish.d. TAe scientific ard. a.tsr+e all, the politiol radliatians of the subi.et e.s.d to dee.cd quick answers. It is rot sarpritinq, thsn, that some of the first repo.-ts did nct oae frvi stvdi.s spscifically d.aiqn.d to .waluat.' the aQpoEt.d aasoeiation. They ware t8e rwlt of cleNerly analyzeC infernation that had bes+ collccZad for otAer purpo.... T.+ese stvdi-s werR vary valuable in thi+oring sam liqht on the degree of the susoer,,.d association. but also they suffer+d frow .Ktl+odnloqic d:a.e.cks that 3.st soK doubt on the reliabil-ty of their resu:ts. Later studies hew provided further .nd snre epreific inforsatim. r128-

-129- Hcaavc. tde acrolwic naturv of t.he pcvblrw. alonq with LY inferfactirn of presert r.r.crrh tools. Mee prscl•.idd tns estaplialnent of definite as+cluniors. As it is almmt alwys um case in epidaniolcgic resesrtfi. it is the infcn+stion frrni many diffesett studies that will pcvvidt the evidenc+e neceasary eittkr to re)ect or to aaoeVt the eYistenoe of the association. 'R~ prexnt study %as carried out to :ur',her assess the relat.icnship be^_wees passive ssokinq and lunq onoes. as well as to pcvwide ansuers to srne of -the setlxodoloqtol peohiems vwaantered by other aLthar:. In dtiis chapter I will present a dise+ssion of the ganeral aspects of the study that sey Iiave had an aff.rt an cur zewltse sueh as smple sise, oomposattion of our stuuy populatio+ (i.e. incluaion of both sexes. inclwion of formcr smdaers. et=. ), approach to data mll!nion, ctcioe of statistical t.c'niti7uea. snd so on. Rhis yill be lclloued by a nore speoi:lc discussion of the tesults of each of our exp-m.ure variables. focvsinq on the ry in which these results compare with others previcusly reyortud. 7Tr present study is to dsts the largest ever oonduct+d on tlus subject. TTr ispliotinn of this fact raflwcu on tix faprwed statistical po+er of our study ovr other studies with regerd to their ability to detect differana.s in risk becwen the exposed and tiir unexpoeed. ttien the sstia.te of ',s expcted to be ccderste or smsil. as has E.en the 4se as suqqastad by previous epidmioloqic ard laboratory studits. omaideration of sa.ple size takes on sp.ciai itRurtanee. Our istiaution of sanple sise :aquu.rcV to detect an ef:c.t oi the order :!por-.ed :r:Seperc:ently by SirrfanN ard 'irichocnulas oSVz8CCzoz -i30- cesulted ln a fiyure of 450 casrcmt:ol pairs, qiven that a setdied design would be used. we were able to asseeble 439 pairs that .rt the eligibility csiteria fully. A sasplc of this size arovidC a a close to 95% probability of oorsectly re}ectitg the mill hypothesis (po.et alculstion based on an odds ratio of 2.0; proport.ion of exposed crntrols to spoue smokurl, p, of 0.25; and aionifianee level, -C, of It has taxn siyqested that the estsmiEis foad by Birsyase and 17ichcpouious are too hiqh, in liqht of rhst is lunorn aGa,t simlce uotake by naR-sadrers (the uptake has t.en es?.iatad to be aquivalartt to the smdtiirg of 0.1 to 1.0, cigarette per day (6SI). It Collovs that the use of such estia.tes wuld ovsresclaste the pa+er of any qiven sample size. An adds ratio of 1.3 is believ.d to reflxt aore accurately the risk associat.d with the level of oposure attained by passive sodcinq (65). 14.~ we use the latter estLnate to calculate the statistical porer of our sasple sixe, tle resulting figure ts still higher tAen the cvw.ntiorlly amepted l06. Rurther.rar.. ,fien in these ealculations we clwiqe the value of the propartion of the exposed controls fcas 0.25 to 0.56 (th peaportion of exposed controls actually ob.arved in our study). cur scatistial porer incz.ases above 9011. Ttx difficulty in qathe:inq a large .naqh rnr.+bei of cases for study derives fras tt.e fact tast most lunq ea:-n.r mses occlr armnq snakers. Even thouqh the latter are pcoba.oly exposed to large amouwtts of envirormertt.l tohsctA s~lce as aconsaqeenre of both t.tieir otin srtokinq ard tleir asscc:aucn vit-I o_ti_ ym;ezs, t.•Wy are u-uuited

for a study of the effects of passive ssokinq, sinae tleir o.n habits aould tend to .ask any effeet due to envirarmental smoke. w recognized that Ve ideal gzoup to be studied w a qza~ of neves Am"+e^s. It tcooe evident, howeNer, that tne population we were workinq with wuld not yield enasph never sadciaq caxM in any practiol period of tisr. Instead of raducinq our saeple size at the espersa of statisticil power. it a.s deaided to include ex-sndcars in the study. The asjor praspecZiw studias (90-91) heve shnwn that the risk of lsnq canaer a.ory foraer ssrieers can deQeas., after 15 to 20 y.ars, to a point rhert it teseables the risk of the never mmlo:ra. 31 the average the .x-amkars included in oua saupie had quit 18 years pcior to thair participstion in the study.. .1re were arare, noryethsless, that wes sesiduwl affect of past ssoking habits eiqht still shor in our data ard thus oamfoud the effeet due to passive snokit+9. . rae :ooic thr.e stepe to enwrl that am cvnclusions wuld not be affected by the inclusion of fosear s+.doers. 4hese are: 1) the design of the stvdy establislyd that a.er and ooir.rois should be setcl+.d on .hs basis of .dcinQ history, 21 .ep.cate estiartes of effect were obtained for sodkers and forner s.nirars, and 71 we ;.o+troll.ed in the analysis for a variable that reprssentad the rnmober of cigarettes the ex-yrokers used to awoke per day. Our rasults prvv:ded evidence that the inclision of forn.r sso!asrs w.s not inap4seQiate. in no tase did +.e ;:nd di![er-ncss in the estimstes of effect l+etween never and forser snckers .hich wert statistically siynificar.t. Nevemheless. a :esidul due to past ino:.iny hanits +.as :oun. ibrmer m+chrxs -152- of I padc of cigarrettas/day shwwd a statistiolly siqnificant odds ratio of 1.38. It should be saphaiiaed thst this etfect was crnt_-ollad for in the analysis (as ez, laind in rnmher 3. ab.Nel arYl therefore is nor to be viewed as respansible for any effect foud for pessive e"kuq. ltaa sewrdua's ham rsTzict.d thair eudir oi passive ssokin9 enClusivsly to eo>Aq. This is aqa.in related to the fact thnt it is easier to titd non-stokin9 ltn9 carws usas stixl 'omen than a.Qq aen. In this cfioicy it is io@lied to a cartain acteist thst results can U.n be extrspolatd to mn. That !s, it is asaumsi that am are as susceptib(e to entrlvnensal amdce as wan, although this is not as easily observed in a mle papulation bcauae am acauirt lu+q caneer as a consequenoe of their orn ssoking. Wile this proposition s+aiaes sense, we felt that it deserved to be eaaluated.. Rfie findinq of an association in eales wuld st.w+qthsn the evidoroe of the carcinoqenie effect of ervirotw.ntal seoke. ota sttdy inaludes also the larqest series of aale na+-efoking lus+q canesr a.aa ( 28.59 of these are never siolaars ard 71.51 ars forsrr senkersl. Our cases were drawn fros sll diagnostic and treatznent facilities operating in the study area. !`srther rnecks for cases were perfocrrd at r.he Nw York State 'itnor Reqis•.ry. Dsapt for a few cases .ho e+y nave left the area in order to ge: w.edioi care in other reqions, we are aonfident of having detected all i.xident cases oclusiM durinq t-he study period. Ouz rtsoonse rate of 761 can be mrsidered qood :cr :met colnventional standarls. we do nct fvve deui:ed in_•orancion on

the rhara^erisuo of tne individunls •R+c refused to parti--4pete, but it is the iupression of the field rrsrarch aasis<ar.ta that the reasr++s for refusal .+ere erze of a nedical roauze - i.M- t-'+e case "°s too ill to participaC' -- rather than assaciated with any particular sociodenroqraoruc c'tiLar.aristio. Oar_ect clissafiotion of •hS study subjecta acoordin4 to tts•iz sacking histery is rnueial in studies on pasaiw seokin9. Inclusion of oersnt sedrnrs .ould sppaiously i.-xxvsss the risk for the sn+dy disease. ;ten snokitq sutw of the reportad caau rs ascrrtau+ed as part of aetssaininq .l.qibility for participation in our study, onlv thcse reportad in the ardinl r.mcds as former s.ek.rs, newer ssdcers or of unlna+n ssekinq status were co[+tactad by ttl.phor+e to Nrtlrr cortirv ttris smdri+q historitl . In additi.on to this, the research assistant rould at tMw tine of the int.rviww further question t.hesa subjects in this r.qud . The ns.d to confia the sddin9 histary stat.d in sodical remeds is Estt illustzst.d by GarfuJcal's findinqs (30). In that study elass to 401 of the arn+en with 1ung cancer classified as non-ssr~ars lor smoki.ag status not stat.d) on the hcspital reoords. had bsn smobar3 at sams tlas. in our sndy no a:-avpt .as nvie to oasfirn the ssokinm history of lhose rapor'_ed as sadurs, or even to establish csstact with the.. St seem less 1i kel}~ , ha+rier, that a forsar or a rr%er smaioer .wld repor _ 1im/hetself as a vrzant wokar, this escaping inclusion inta the ataaw. thc cioee scvtu+y of ouc eases .cd crntrols raqarCin9 t.'+a= i I sncadng hutories sl=mqly irsr.cates that aieclassificat:on .as not a likely ev.nt. Previous studiea have Cem+ queati.oned in this regard. 1he telativs:y hiqh lunq canoer aoctality rr?+4 nan-smdetn9 women r, osperad to ssokinq ao..n in Htray.s 's sauCy (25). aa we11 as the high mortility for other ta*aooo-r.Ltad dis.as.s. have suqqestad that srne ..ok.lnq - say hsve reportd sHSSMsiurs to belong to the rnrr s.oltinq yroup. 'Lhis wuld rot be surptising in a traditional .ociety srh as Japan. 'Cw corsaqusoos of such sisclassifiatLon would be to over.scisata ths aff.ct dus to passive smokinq. I}w Gr.ek study (27) can also be eriticissd an the same Qrouyis. The clais that sion th.n 750 (40 out of 311 of lusnq or+oer ossss .rre diagos.d amsrl oar--okinq - ia incarist.nt with the pcoportion reprxted in the lit.rntur... It is aouidsr.d tha only about 10% of all iuy eanar in r'mn ooum in rbt-s.olas.."s '(ths mrr.sponiir+4 fiqurn for seo is 2.0%). It is therefon liloely that this study cvsfousdd the effect of passive ssokinq with th effect dus to direct ciqarette srokirg, at least to a osrtain degree. 8efors our sady, only GarfiNul u+d cowcrk.rs (30) Md carried out an uideperuient review of the histoloqic diagnoses of as.s. !n sost other reyarts the aut.bn had relied on the disdxcys diarg+os.s reoordsd in the p.tirK's nediol r.eoeds. '1he d.gtre to rhich thc.+se diagnoses were based on •xam+nation of p.tAoloqy spscimens is varied. in 1Yichc.pouiow ' study only 35% had an histoloqic diaqrnsis (27). The crorrespondinq figure fa Akiba's stidy. is S71 (32). Pelayo c.orrs reportt 97% /29). Rxh Hlraya.et's wrk and Carfin)oel's first stadY

-13s- usiny _he hr,ericsn ;arwat faoety data reliad exclusively on death csrrtifivtes (24-25). 'Cr ispnrtanct of histologic di+rynoses is that it prwides the eo.t amustu wswiresfnt of disease status. Inclusion of oth.r onoers can biaa the results either ary from or toracds the null valus, deVeedln9 on w+ether or :wt they are related to asokuq. the latter is the sest likely situstion siixx irrlusion of ocber cancers wuld tend to be random. An additional adventa°e of obtaininq the histologic diaqrnais is that it allows us to estir.te the effect for spszific Aistoloqic types. 'fhis is of partia+lar inporcance to uncoqer any specificity that would streryhcen the e.Irlerns of a eausal rela•_iauhia . Ma rare able to abuin alsost a 1004 of histologic oontinotion. aid given the emoellent aqzemert brt.+esn oia reviwi.r and the initial AosPital diagiases w are certain that sdxlassifiotion of disease status did not play a role in our find.ings. Mother ares of eQ+osrn in sadies of passive snokinq refers to the eorrer classification of stssly subjects with regard to e4osuce. In the present study. as in all others that have been published, the study subjec.ts wsre :lasaifiad busd en the infonstion on esposure obtained throuqh intsrview. It is advovl.iqed that this approsch is less than ideal. At the en!snt. however, it is the only feasible option. It has been sugqesttl (u an alCar.rtive to interviowl :.'a: laboratorv sessuranuits of cotinins in sa;iva or blood cot1d be used to dete:mur esposure. :Tese would solve the prvble+e only partially, because sudi tecYVUques are u.tiit to measure expasure that has eccued t:uctnut a lifet!ee - a messuremen! [.ti+t is :ncesar,., vhen studyinq EeV%RCCzoz -136- a disease thst probably oo0--s as a result of a prolonged sTos<y-e. Nevertheless. laborstory tecistiiquas can be used to validate and refine quastionnairEs that inqui.re about e-Vosure. The ;robiev of recall pused by the use of questionnaaas is sagnified s+e, the study subjects tlmrelves are not a,.ulable to provide the neassary inforsatim. In our study wc heC to resort to the ufe of surzoq.te respQdsnts in 33.0 1 of our interviews. In order to aini.tiue pcnblea a[lsinq frm this approsch we satdrd cases and aonteols on type of interview oonduct.d, and obtain.d separate estisetes of affect for self-respondrtta and sutmqatu respmdents. 'ltie results. with only ons esosption (estia.tas of effe,R for nuaber of years smoked by t'w spoursl, did not differ a: aooowt of the individu.l .M had psarided the infarwtion. Similasly, Cfctinkal ()0) -.who only used 12s of direct intaavlwm - did not find significant differenaes assnq the different types o' infccners. Ehviron*ental tob.oao sneic. has b..n for the p.st decades an ubiquituous pollutant to .ttich everybody is believed to have been -gosad in different deq.r.es. .Arther )oo,rinqly or urrmoidnoly. It has been suwssted, therefate. that all studies on passive soking suffer fros sisclassifiation of exposure since all claim to havr identified individuais with zero exposure (that is, the individuais used as the non-wjqosed referent qroup). It is ispossible to ascertain t.'x daqrae to which sueh a probls wy be present in our data. as it wuld bs ispossi5le to determins Mwethar or not sucA apcoblm - if rresent -- affected cases and eantrols equslly. In the ev.fnt of ~..M.-----.

-137- differ.ncial Wsclassificatson in osa study, ae w+ld axpect to find bias eit.'+.r away fros ac towacds ths nu:l v.lus. If. an the other hud, wxliffsrst+tial sisclassifiotion atfaetarl our r..ults, the bias oauld only be an attarnation of the estU.t.s of rffact. Rx sum contln»usly s,..suzd variables as the oaa as.sssd in this stody, the ..gnitude of this bias wuld depend on the wgniandr of the correlation bttwn trr tius valu.s for tsposurs ud the values of exposaae as s.+ssur.d. us did oot abtiin infoewtioa an caetain variablas knw+ to affect a+4osur:. Vir+tilation, sisa of :mis. proaisdty to the saokar aed other ecrditiau tsd.r whieh ssdclnq taias plaee influeros the aciusl dosa r.oaiv.d by the pnsivt ssdrar. 14 aclcwl.dga the diffiouity in otxaininq data in tRis raqacd. especially if irSocs.tion a6out a lifetln is bsn souqht. we also bslaiasd that askinq fnr wch e.tsnsi.a dstail in a quutioln.isy wight pr'aV.nt our aoquisition of sun basic irSocration, by distracting the iat.rvisis froa essentials. Naw t.hse factors s.y have influ.na.d ovf results can not finally be d.tm.insd. A11 Our controls c.r fmm the files of the INSw, York 8tata Oepattirflt Of Matoc VrhlClaY. 11e havl.data to support that they are caparabls to the nsra in the sost relevant eharseiaristie.s. 'thsre is. hm.eMer, ons pote+t.ial souros o[ problsss. It+srss r arn oartain, b,., definition. that all cont=ols wcs liosns.E drivers. w do not knov the nses' starJs .dth rsqacd to dri•Ary. It d:•irinq atatus ia soae`ev related to the cisk of l~sq canoer aithrr diractly oc, euch aore lik.iy, ;nC±_-ee.Iy. our results sLqAt havw b..n suCi.et to hias. Ib+ryer, if any possible dilfa=.nws in driving status r.fiaR only differ.no.s in sociosaaiomdc sutus, our eontrollinq in the analysis by 1tc17W should hww taken c.ra of the peobh. Of all the osanss.s ais.d to assws eVooas to passiw .nokiiq, wi. .rwrsei+t of the apaw ssokinq hsbits has been the snst frequently os.d. 'fhis s.y aocoint.' at least partia/ly, for tns f.et that this is also the ss.un most often reported to aha+ a posiaw associataon with lug osrcar risic.. Asoiq the diffsnnt yys in .fiich spvuse s•okinq has b..+ quantifs.d. the e~ of ciqaratt.s .nbd pr day .lp.ars sn.t often in tMs litusturs. Maesr of p.dks a year saoked, total nabsr of clqacatt.s sdrad in a lifetisis, erabsr of y..rs smokinq, and euEsr of dqae.ttr ssnlad at hone h.vw also bsn usad as a p:ouy for ssposise to pasaive ssdcinq. Sn the pcs.<rt study re hav+ chosmn three diff.raet w.wasertts of rsposur.s 1) nabsr of eiqarett.u sosoksd a day by the spows.. 2) esabsF of y..rs the .pou.e ssakd. aed 1) total msflasr of eiqarstoss sdod .hile marsi.d ard livinq toqetlrr. w bsliaive tlwe thrs ~sur~.nts ooipcis. inforwtta+ on the tw sast isparta+t diarricns of .aa,osia., intensity ard duraticn. Additiosaly they also pciwids results easily oorpYrabla to thOsf in other stlditl. Pr.vious studitl harw shOwn that nort-ssokinq wnn sarr-itl to sso;aers of 20 or were cigarettes a day sodhibit a risk of developing l:nq ar+csr that is approKdsrtsly doupls tto risk of tho.e s+.rtiad to aon-snolcers 125.27,701. 'lhs .stls+tas obtai+rd in thcse studi.s hsw bwn reported to be siqnificarct, ard .frn couidsr.d alonq with the

-139- -u0- e.rima:s tor ot.hcr levels uf er+ocure, t.ave alsn cr.mn [ossd to show a dose-respQtise re:ationship, whics further suppoas the case for a Our resulto regardiig ths.s•variable cwsit be aocaxsnted for by causal relationa,hip. ibt a wmerh+t oaparable lEw.l of ejVot~¢a t+e the L,clarion of fao+nr ssokars sinoe any affect due to this .ere sssable to dttact any significant iuwsesse in lunq eanoer risk. cirnstanw wuld act in the oppssita directioe. that is, wuid 0.r point estis.te of effect foc theee inEividuals sarrieL' to ndcers inflate the tiLk a-ibotabla to e2poeors. tbr can it be eaQ+laiird by of 21-40 ciqarattes/day %as an adds ratio of 0.906 (95a Qs .609- the use of surroqau interriews either, sirne no significant when esposure w an.lysed as a wteqarical sa;`..able. Fbr 1.35) diffennaes wue foisd het.+en the eetLrtes foe the latter ad the , iniqher levels of espoase N foisd risks aboue the null value, but again these findings were not statistically siqnifinnt. T!he analysis of aspnaas data as a castinsous variable ras ecsrist.nt with the estis.tes for aelf-te.pasdsts. '1h. effect oould not have b..n diluted by the inclusian of wles. sinae they seem to stnr a sissilar r.sQan.e to that of Mo,en. linally, uruqual distributi.as in easea and oontzols a6ooe a+altsi inde.d the ef`ect fousd for a diffetess•±=1 of a>qassae of 20 ciqaretees par day me harely abwn the tull value (OR w 1.011). r.aardinq oon:ouu'Snq variables oanneit be reepoirible for the findLiqs. as sulgeesud by the eatnor ahsrqa the estiat.es undergo .4rn with cotifidnoe liedts that include suds a figure. Nor tMre w any thoee variables are included in the analysis. evidenee of a lis.ar trard. Me did rot find an effect dep.dant on the dontiae of aposure 'ehis r.part is not the only onm that has Esm+ wble to aanfiai a tb passive ssnkLnq as aersurtl by the eawhar of yeus the spnwe statistically significant asaociatlcss Est+nen lasiser of cigarettes a~oAsed vh[le ssried aed livioq toqether. 'f~o otlrr etudias h.d s.okad by the spaw aed lsssq ear+aer risks. akida, et.. a1.132) evaluated the effect attributable to this ssasucr of esposure. reported an odds ratio of 1.31 (909 Q.s 0.40-3.47) for those e>posad Oalager st. al. (33) r.part.d eas significant ine;.ases in risk due to to spouse anokinq o! 20-40 ciqaretus/day (reported as 140-279 ciqarettes/reek). Fbr a siadlar lswl of esqawre. Oelaqet at. al. (33) fourd an odis ratio of 1.S (9Sl . Q.t 0.a-2.7). it+ese latter estinates, althaugh not aignificant, ars within the rsrqa of the assprct.d esqntt-Yit of effect. & closer fiqure to oss trsults, ho~aver, res reported by Gariiniaei in his an.lyas of the a.erican Canosr Society Study data (24). fbr esposnre to spous' surking of 20 or mre c3qarettes!dey he fovd an odds ratio of 1.10. Sluilarly his results ..ers not siqruficant. ~gfizBEC%4% spouse ssokinq of 1-20 years (OR a 1.73, 951 d: .52-5.42). 21-30 years (Cqt - 1.78, 951 Q..: ..60-5..10), ad >30 years (Oll a 1.24. 9S% G,: .42-3.53). 'tfrse figures did not sAov aviderrce of a litrar trerd. Nci6a, et. ai. (32) reported odds rstios of 2.1 (90% CLe 1.0 - 4.7) for t,i+ose married to saokars of 1-19 years, of 1.5 (90% 2: 0.8 - 2.71 tor tAose e>qosed during 20-39 years, and an adds ratio of 1.3 (901 Q.: 0.7 - 2. 5) for those eto 40 or soce years. ihese ron significant effects did not shov evidenoo of a line.r 5tend. Carfiuvice:, at. a1 inco.-porat..d a dif!ereM eeasure of duration of

e3posura. 2lsey estj+stad rhe er;ect asaociaaed v_1s aQnwre due to hustard 's saeking in the last 5 and tne last 25 years. 7Trse .aasures of epowsrs ..ce analyxd for ditfer.rn aqe q=ap, histoloqic types. type of respmdust, asd sociaecviaiic sKatus. Qsly arr of 30 estisetes abtaisrd sha+ed statistical siqnifionoe IaFt - . 2.54 for asibess of the los+er sdddie class e>tpoaed to spouse srduag in tne last 5 years). 'lTse evalssatios of our third sranuant of spouse s.dcisy - total =asber of dqsrfttss a1ed dnrinq s.asial life. - dor not saspport an assoaiatian with ltsK ouos= risk. Ztse estis.ts associated with a di:farzntial of espowsre of 200.000 ciqarettes w foustid to te 1.019 f9S9 Qs o.a77 - 1.1!!. -.'LichoFvulau et. al. (27) used the sa.e e.ansramst to asusss pasaiti. .ddnq. Their analysis, howtnrar, resulted in estis.tM that are relatively high (a three-fold i.,Re.u in risk for espomsre to 200.000 - s00.000 ciqarsttesl. Qsd which aiso sisw svidvs~ of a linear trand. 'lfieir fisdinys persisted after costzollinq for sos ispoctant variables. In our stsa2y the inclusion of oosifoudinq variables did swt pcoduoe any changes that wuid rwerse out cosclasion o[ a non siqsifiont associatioo. In rJse p.raqraFhs abvse we h.w alled attant.ian to t.he fact that factors related to the Jesiqn of the stuuy, or to Ue presewDa of canfcasdirq variables. oasld nos tsplain the differances 'aetuwen our rerslts assd those .srKAs>s a oositive association vith soause snokinq. Other circvrsunoes. ha+rwr, s.y have affeetad our results. Srnr of thex aze prexnteA as follors. seVzRCCzoz -142- Critics of the first tw eptdesnoloqic stisdies an passive wrdcun raised the conoarn that the results nrritl out in ts+o foreign couariea (i.e. Japan asd Gra.cs) sdght rot apply to the U.S. populatian. Dposure to sQasss ssdcin4, it .as suqqested, wuld rnt be as qood a pcoxy, aassurs for oy.rall esqsanQe to s+.olas in thU cantry as it wuld be in those .are trsditiosal societ.ies. In those sodeties spaw smdcieg - it is belaivsd - not only carelatei well with overall esqosure, but actually oos4sisss a laqe praportion of it. A raeest study in the U.S. suqqesrs a differant picture for this oountry . lriad.an at. al. (92) have fousd that determining an individuas espossr+ on the basis of the spxse ssoklnq habits say lead to gross erzvrs. For issstanas. they reported that 47% of womsn ud 390 of sen oarriee to s.daus reparted se:o Aars of espaeure at isons. Conuersely. 40.50 of wian ard 19.29 of sen ..rried to flQYfRCkars repOrted 10e e7p[itSe to the s11D1ae of other per!?1s. Itse degree in rhids a sisdlar pavblem sy apply to our data. or the .my in which (if pcosant) it asqr have affected our ossa asd carrtrols dif:erently, cunot be ascsrtadn.d throuqh the available infoc.ation. 4tws, such a possibility r..ins a potential sourca of erroc. Oifferent life oaditiofu in the U.S.. ary also as+ke tir u.e of spaise sodcisq a bed ptvry for espaw¢e to passive s+iokinq. Fbr ia+tenci, the average Jap.n.se hoas is nsdi ss.ller th.n the Am-rion !sam. 9%o indiatocs of hasse size u.d by the United laations illustrstes this 03) r 1I tf+e average sise, rspF.s.+ted by the sran ssumtsr of psrsons par isas.selpid, is 3.2 in Japan, .fmreas in the U.S. is 2.7, asd 21 1+e peetsstaoe of hoisinq ussits vltIN ons roorm only is

S\ in Japan arsi 1..56 in this couniry. As sentioned in anotlrr secuon of this rapoct. rom sise, eoqether with a clo.ely rel.ted variable - i.e. psoci.ity to the wota; -grea'_ly influsnce the exposure dose actu.lly reorlvad by the passive seoker. It aiqht well be acyusd that. sinoe soas seosnt Amarican studaas have supporzed the findis+gs of Llirayaws and Thrid+q~oulou.. objsctiass such u thor raised sbose h.ve already be.n disoarted.. It shoula 6e esphasized. however.' thst only the stsdy by frrfnlal et. al. has sfe.n siqnificy+t findiiqa !a the ess~t.~R of reaescFh without wjor .ethololoqial prvblens. M additlorrl study (Aslayo Oorsas, et. al) presetrt.C qusstirnn.ble results, p+rtiwlarly dus to the ssll asber of sArjects studied. FUrrlrasae the ot recsnt psper publish.d in this oantiy .as wble to urzxv.r any statistically significant association (Oslaqar. et. at.). 'fhe ef f eet dus to spouse ssokinq in oir stydy subjects sny be so small as to be urdst.etable udsr the pmssot mrditions of study. 9xl+ effect, howewr, sdqht v.ll De~*aw detectable if exposure troe differ.nt sourees ware to ace,aulats. 'ffie fact tFrt ue foua+d an effect dus ta espoa:re in the hsuser+old - vhidi includes the spet n+'s ssokinq- srsayly suggests this possibility. Sardler's firrdi;-ws fustAer suppoR it (37). He foud an iherese in risk with the increasing nueber of haasehold ans%bers rho amoctl. trnaodinq the spms+e- lntuituon, toqether with these results, ..ould seen to xy^q"t thst the woet appsvpiate +.y to evsluate 9-As effect of passi.. .o" 4s``%SC.(.C.oz I wuld be through tSr ue of a cuaul.tive iidez of exposuae. Illtlouqh in tMeory this is desirable. ttn esiqssi.es of collecting the infas.tiu+ my praserlbe it. itiaple's pereeption of the =Wanq thst qoes en arosd tAr ry rary L~ onm cimnatanos to another. WuLvass people wy be able to provide aore or less detailtl and amaste infacsaion on ttr aenkiaq of the spotse or other esmbrss of the camily, they niqht not be as precise rhen it aaes to the smokirq of others or to the sndcinq that oazrs in public plaoes. Caduninq suoh infors.tun in a single indoc could reault in the aixing of 'good" quality infocsation with scm 'h.d' quality infocsation, thuw d:ainishinq tto peab.bilitiu of fihdinq qoad estisators of the effect. Lstly, an esplanation for our nyatiw findinqs my ce attributable to ttr fact tht siokinq by the spune is an exposure that is inuodrosd late in life ard sry not be p=olongad ersxs2h to pivdure an e1feR in .nd of itself. Such expcsure, Moawesr, oauld becvr ispartant in the pr.s.ncs of a previous exposure history (i.e. parenEal ss..kirq) or. as oplainsd abow, in oonjunction with expowae fro> other souroes. w foud an inQem+e in risk for lunq canoer for tbose ejpasad to the sedee of ot.he!rs at haa. 'Itie odds ratio assoeiated with LSO person/years of t%ppsure ,,.s est{msu4 to be 1.86 (!36 Q.t 1.22 - 2.671. Only 3 few ?revious studies haw inchaied a wssurm.r.t of exposure at hovw as a ask factor to be er.luated. In such sa-dies, horevar. ttie seasureeerr_ ot exposure has bean rudia.ntiry, ussally

-145- sxpressad as a dicot.9o9 (y<rho aoqoelusl. r+e believe our saasursavt to be sore aosplr_e 9is.n that it ocopr-ses a prary fer the iatsrssity of the espoaire (sswbes of s.dkass in the hoisahold), as vll as for its ditr.tion I:x.oer of years of oohsbi*atioa with s.chaars). St can also be regarded as a o:.pcahn+sivs sswre. a±'xi it includes all household .a- s=s for t» subi.cv' z+asidr +oe hLstor (ths latter osp.ndin9 for the scst part of t'aeis adult lives). 'iAe collection of su3l ectrrive inforswtton. Isawwr. w not vithout its pevbl.mt. as evidsrKad by the ocvrrarsa of eosm missing values. we have explainad the rys in vhich ae appsoadrd sudh Frvbles Iry ispatation of aisainq wlur under tw differsnt asw.ptians) ard feel oQSfidat ttrt our resilts haw not bs.n substantially affected by it. Ttw tollowinq ars the results frv other studies ass.sal+4 the role of eq:osue to passive wddnq at Ao.a. C'Arfinkal et. a1 (I0) cLssifid the atpowue to amdra at hoes acoocding to hov rawK .as the expowu+s. !br thoae esfneed at hone in tAS last five years (ysi(eo sxposae) an odda ratio -2 -`ound for lwq ou+oer of 1.22 (9S1 C.c 0,92 - 1.62). M awn louar sstiisate w.s rspcet.d for esposure in the la.c 25 years (at a 1.1S, 951 C3..s 0.l9- 1.19). 'Etree results r.re aot statisticslly siqnific.arr_, as sho++ by the astfide~os intervals. A aare Wined seasureRent of expasure at arna %as also assessed by firtinisl, but it was lisnted to spcusa ssoki:+q. A significant linear association .as :owd fc.r sx*xsure tc the spouse arokinq at hmw (no association rs fme+d for the sw,c:.r? that occurs ourside the hnae/. fudler et. al. (]7 1 reooet.d an elewted risk for tlror expwed at h=. '>re odds ritioe associa•ed Sstz&CCzo2: -116- vitb 1. 2. and 3 ar wore aiukinq Nousehold .rbers .are 1.5. 2.3 and 2.0. tnepectively. Lven t.tsxslh only the last fiqure was in itself sta•+s*icarly, significtnt, the threa figures uqhether follow a statistically siqni!ica+t linear trad. Rabat ard byrdar (34 ) fousd no association bet++.en expouus at hoss+ (yes/no exqwsure) ard risk for ltnq crsoar. '8rir ..tisaes, ho.ewr, are b4sad on a study that includad only 25 luq aroer patients. ln additi,on to the osera!1 ef_`eat due to hous.hold aqouue found in oia data, thars is one ottsr finding that deseeves speci.al oonsideration. The effect 'of passive s.oklnq in the has.lnld w fousnd to be siqnifieantly larger for .pid.mLd/s.all oall histologic type than f= ads+oc:Lnona/ar~er oslls histologic type (wd odds r.tios of 2.67 and 1.62, respectively). Cantrolling for aontourliny varialsles resulted in estinatr. that altlough not siqnificant, aers nQrtls.less aaristent with cAe abow findings (adjust.d al - 2.32, ard 1.l6. re.peexiMly). As explaitrd in th+ paraqrapha abose. only a fev studies hevm classified th.ir cases in reqard to listoloqic diaqnosis, and ewn feuer havm included a large enayh nuk,er of 'cases of ead+ histologic type as to be able to sxaaine spncific associatioro. 'it» notion that passive swokinq s.y te siore strongly asaocia-Pd with epidezweid turors. therefore. conas nsinly fran studies on ,~e effscts of direct sodcinq. Althouqh .aim of these studies have detesannrd that all histologic types are related to ssokinq (see Cu3ter 1: '9idc.iaq ai+d Histoloqia lype of 4nq ;,anoer'), sost of te+cs srss to a? r.e that the strongest relationship is observed :or :p.leso.d vA sTa11 ^el! turors. t`+e suqqasci.n that

such fif+dinqs hold as well for pssauve smokinq pr©viaes upport to ih• hypott.sis ttut emriranental woke acts in a sinilar r.y to .sn,tr.am s~ and th.% sumw that the only differar,os 5etween the tro types of espowtss ara quantitative. The one other study that fourd a statistically significant higher risk for •pideaooid type ..s that of C+rfinkrsl's at. al. (30). 'Rty raporC•d an adds rztlo of 5.0 (9S1 Q.: 1.43 - 20.18) associated with the huab~srd 's ssnk.inq at hoae. Another study repa:ted siaii!ar result, althouqh non aiqniFicant. Oalaqer et. al.(33) teForted an adjusted odds ratio of 2.88 (951 C.: 0.91-9.101 for epidersnid/sarll osll carcincoas. and aa odds ratio of 1.02 (951 aa .33-3.16) for admnra:einoss. Zhe iaportanas of psuive ssflkiip in the houstald sust be put into p•rapsrtive with ryud to the avsK of espoWe nsaessary to oe.•r.e an effect of the sa7nitude describsd. w have pres.nc.d the eseisst•s of effects as•oriated ait.h a.I exposure of 150 p•raarVyaars. This is aquivalant to h.vinq lived over a priod of 30 years with fiva ssolars, or with appeaais.t•ly 4 sookor• for 37 years.. :t should be eiplrsised that sudf lwel of exposure is rather high and s•y be unoorman in the gvnessl prtpulat:an. In ooF study 9rsxip, for insea+c.. valy 7.56 of the cases ard 4.31 of the ounrsola attained that or higher lwels of aqowr•. Indsd, lower lss.ls of asposurr. are sarh sort tvman. fhr exsrple, 251 of the wses and 20.51 of the centroly repoftad e7qCt11Ces 5ltMe1 50 and ?S..• l7lrlQS/yoars (eQllivaleflt to (yYll1q lived ZS years with 2 or 3 sot7ker3). 9uC.h levels of exQ07LLiC, - hwever. carry risks that are Irore mCdest (QR !or S0 FersO<1/years 1.23. for 7S person years - 1.36). I -i s- 'lhe e~osure to passive ssokitg that occvrs :n t.he vorkplaci has bs.n thour,ht to be .utensive a+as)h to d.serve oensideratian as a pot.ntia1 risk factor for lung cancer. Irboratory data has stown that the urinary ootinin• levels of nor-saasrs wrkinq where saok.rs a.-e pcesent are significantly higtnr Wm tMsa of nm-wok.rs working with norrsiuiaers (S6). The epidesioloqic •vidw-ae has not been yet repoctaf. CArfink.l fond sdds at.io• of 0.88 (9S6 CL: 0.66-1.18) and 0.93 (951 Q.: 0.73 - 1.18) for sadu exposure at .ak in tlr last 5 and 25 years, respectively (yes/no esposur elasaiiication). Rahst and Wyrder found a b.rely statistically siqnifiosnt aff•rt due to exposure at .atr (ys/en expa.urr) for wles, Oot not for fearl•s. In their study 18 of IS cases repartsd to have hrn espus.d to caqarettvr .mke at uork oanpLred to 11 of 25 m+ttols (Ot - 3.27). Io aur study the risk associat•d with exposure in tht wodplaw .as statistically indist.inqudshable frow ths null value (Ql - .9740, 9St CSs 0.974 - 1.0141 for 150 prrsoNyaars of exposur•). 11s also foisd that spow+r• in the wrlq+laos asy noE be as eetrrive as b.laiv.d. 1Marly a third of otisr cases and oontrols repact.d no exposure at all in ths .afplaa., ard alanrt 500 r.porttl espowu.s of lsss than 23 psrson/ years. this distribution results in a sm.ll nster of highly e~ar.i individua's in ahich the effect, alt.houqh more liksly to develop. would also be PDre difficult to deteCT. T'iJMlly. aGRlRllat.loll of esposun in the housslold aed espawue in the wrfplao in • single ~o1rulHt of exposure (botll Mlre Oasa•d in the •aes 1lRits) provided non signifiCant results. We helilvtl, as 11lAtionsd earlier in this ehspter, tr+t this is the result of owbinuq two xaasur.eenta .nth two di:'a:•r+: degrees oE s+snsit:vity. Mureas ~cJc:aq by a-bezs

-149- d d b h h i h If we asassne first that the relationship found Oet+ween social ose reeeive y t e elac.et wit of the famiiy sny be closely eor t e cavxar is tzva .~e easured) atd lun i +nkin (as he bs e Y+rs' ssoki rd t t l i -w h g . . ve s q espasure to psss , rq a nn n cv r passim s+ro Rer. t e aar. twa at dose receiv.d by the passive asdrar msy not he as close. re11 asiqht .mder wAst oould have caused lung ceancxr paumts to have ber+ esposed to less social passive swokinq than were the 'he.lthy' Eefon ours hd fecluded the as.esment of )b other stid oontrols. Any respEans. alarr thoso lines wuld be serely sp.wlative; y Suppose that the oxzr'red to us vie one that ha th foll i b .3paatae to passiw auking in sacial cit+rlstaror. we ladc, . q s s o ut e uecioe of espaure ed the dsra+e as a ootis. d.ed d velo s h ve d t b h i Q q in ase p e a erxice y t to oompata our urdwpecrt.e therefore. any paint of re c the `=aasal• aposur. lts id A adin to our ru iv results. w fousi a d.a+ase in riak tor lung osrv.r with ina.ased i (OR - ad t fl i .d i d l i , q u nq. a e mo to pass •ould be tne ane derived tzram ssntr.ing in tir has.holdr that is, frtw pass ve . ex c-ts o n ect eoc . c ig exposurv to an t~ It oDUld ificant results id si o os1ge for which re fo l eM th 0 639 7lS) Given all the followinq - 11 t!r 951 Qa 0 520 - 0 . u qn n y e P . , . . . aaa:e vtuch ooavrs in the h®e carv+ot be a - d that ths es b th v h I t rt t h i i h n . 4ie p e en a e s at ~strwn eqosure o pievLOaw rapo s t w an assoc an a rass ex o ur social d) rh i id b sil v id d ( assive ssokin canctr oct u h a 2) our ox+ re.ults that d lun p w , t s n avo a or, csmot e ea y a o e p e q r . upp s g i i l has a osrtain .A fi the irdividu t3 i i m r+latio tu lat l.st for one of the es l) the owr* vriables) c resistaroes s a e nq over t sa c f ro p p , h di th h b dd bli d h t i i deqr+ss of contzol. If those .fw are to develop lung canecr develop stu .ve esu s e as.oe an rewa.n act ve m ng es at e t at l i t th t l b th f i ard 4) the l.bucsto ard the diasesse dats that lrs oentitsled the ung y ac a s svgqat.d e ratory syapxass (as y rssp ear ry , the tor function (671) d d b a l i t esicir» nie c ies of tahseoo aokt - the of a s ossibilit y y , tar. ptecl. s y poor rwp onar s o rs ge p op rt p y amtzadic- effe lsks s o t of assive s*okin oe lun c.na em s.y choose to avoid aQossse to ssoks in social cira>*stanoas, but .: ry c q . p g r r f d d fird it harder to do so at ho... :lws, oqrosure in t.w household un ov e . wuld corxinus to permit the disaase-g.neration proc.ss to be In seusitiinq for an esplanstion of this result. Me wut muider tw possibilitiess 1) that the assodatJan fosd reflectx a true _siatiorship, ard 21 the: tls tesults ars pcoduct of data artifacts. A third ;ossLbili'_y. t.'Y role of cnrnee. has larqely bow diseumted thraigti the appliation of tests far rhich the staastial siqnifianoe has been set at O:OS (i.e. results sueh as the m+a observed ray have cce:red as a consequencs of clancs vnly in S out of 100 'triala'I. osV'CVAgCCzoz -cMletd. .Asile at the sw t:me exposare from other souac:s wuld be diminisned. Similarly, it oouid be peaposd that, in a,dditia+ to passive wnking in the Boisehold. there is an as ywt unidentified risk factor cesponsiale toc a oariderablw propaction ef the aaes. Inaiasirq expouue to such a risk factar aaanq the cases, horeser, wu:d also h.ve to be associated with a d.asssinq espanae to social passiwe siokinq. In this ..y the raverse associatSon observed between social passive sAOkinq aid lung anoer wuld ye real, but by no swns

-151- wsld be cnral. A a.cvd poaaibility N have eansidared is that our results reyacdinq social expowae are artitaeuul. ltut is .oat liloely rara onsible for this is the use of an aititieial irdac of exposuae not pcwionly validat.d thtoqh ©rcpar aathodoloqic studlaa. Soat facroes vhid a~sy Asve affaatea the quality of the intotrtion ooll.cird theouqh w.ch smotarli.as as follor.. Pirst of all. tha psriod of tiar oa.rad by the qustJoatiise itsss is very lwiqtly. tlidtinq good Woasiaq on day-to-day .r.rts in the past rsluiiss a ooaaidatabl. effort of reoollection. l.rtitularly tar the first y+ars cf lire. a ocnaidarable raaolleition of the lrr.+t's or the ta.ily•s sociai life, Oic.i are the aan vaAiela of axpo.ure to enviza..nr.al ss+oia. is requi;.d. Sec~oedly, the infornation as it Ms aollertad did not wke any att.apE to sa.naY tntaialty of .xpoe<e., oor did it seabliaR the relative ispoiT,anoe of esyowne dirinq onm activity as app~..d to ar.ot.tiar. B+era is a rrl quaatian. for inatara:. it visiting frierds or relatives rho ssoM,e 3orn.ya the sass dar,na of elqosure u wwid be tncvrred by going to Ears and r.strltents. 'Qws, aspo.nr-s for a11 activities ws waiqtn.d equally. 1!hL-dly, the speific soeial aetivities spacifid to ttn intervier.e ia order to elicit a rsspassa - that is. t.hone rePres.ntative acslvities fca.o to be cv+d~civ. to exposur. -.,ere .'ery l is¢tad (900 sp4..r `+~' x:.. 9sqs 11] 1. !or srss-v st,udy sub)ects ruo -152- ssqxosare might oriqinate fro. nery differant sorts oC accial circ.aW--vrces cltir: ttnst esFlidtly qussria+ai. The fa! 1're specafially to nettim t1YSr artivitias wy omivably have pcodWsd fon! u!dlrrepDrtinq. flvurth, the ust of tw diifernt units of time to seruure frequency of oqroai.-e wy also have aonfua.d the respuda+ts. 'C.e less =requent wpoauaa had to oa reporttl aonordinq to the nurber of tims a oonth the expowse ooaa.d. vhrr..a the aost fsaquent exprsures had to be reportad on a par/.w.k basis. rinally mqxrelwrioe of the infors,tlon bainq z.questd s.y have baan inpai.rod by t.ia oosplex nat~e of the qunt.iow. Fich inqui.cy dea.nded responaes ahieh si.ultarrously ayntl+atisad lnfaoatior. on three dif f ererr: aapects of es<owue. I typs of social activity,. ag" of exposuus. ard fr.que+r'I of ssposurel. 8uch quastior. piacsd alssoat at th. end of a orr hau int.rviw. way have ba.n arsrwrad vithout the =noaw-sation traessary to pearid..ltoqather adsquate resporses. w M.ve not srntion in the abova list the use qf stiu-u~ete re.pudants as a poasible souraa of data artifacts. tfr reason is that we do not have any .vidsnce to wpport that the estimats N+a'-ned !or self-respud.nts are any diifarant fro' tAo.a of surcrqate respanderr-s (This it a situation that is also tsva for all of em exposure variaAles, with on s~ocsption: the rneOer of years sdo.d by the apousel. eut qiven the cvmom cv+cvxn an that sa.tar, we telt that it should be wadc esplicit.

Isperfect:ou of the 4a;:a ooliectan tools alar, horov-r. mpy wc be snffident to esplain our sesults. it is the .ay in which sucn isprr:ections a!fe:t.d the rerpmsse of xses vis-a-vis We respaue of cvatxols trat msy provide additi=.sl urderscandilg. Aqaiu, we c!.n only specu:atR on this srttes. Our nisui:a suqqest that the ars oay havt urdereportied their e=as•Ye to social pWiwe saukinq. Snch a prcblem as oftan .mntioned in reiat;,cri to w---rlies ia .dsich the suspewted exparxre .u socially umcesirable. as is tlw ax with drug or alootfol use. in the Frilff+t i.'1r.at1Cl. !laevar, there is no wd1 stiaeetization attadhed to ripoaiure. lt is, attAr a;l, the ssolcnq of others, and rnt the pstient-s own risk fastnr that :s being isplioted. never',hless. patients for .fttevar reawn s.y have fel: the roed to miaimise theL- e:Tosure to anything that potentially they mi,ht have been able to avoid. (nuureportuK of aWwure by cUntrvls a.y aiso, in theory. have pcoduced the results otrerved. in practice. tia+ever, there is no cam:rl'ainq reason to beleiv. that controls .ould - - bsefit fros su¢h behsviour. As a aoRsequence we have disoounted ovareportinq as a possibility). Throuot+wt this doaaent «e h.w presented a sumM.-y of tt& infon.:ion tjz--en:ly availav!e on Lhe relationship betti+een luir, ~.arwer ud passive sickinq. Li !hs !:zst oar_ we 'sava revi-wed tGx ws;or epidenioloSic stuoies on the suc7ect, as rall ts sare relerant lacoratcry data. The =joc por•Son of this des-asr+t, hor.ver, nss :.cen devxed to t.'w des.•-aption of trn .a,tiodoloqy ard the results o! this. iarqes! epidernioloqi^ s:uc'v so far :_r,&jcced to -;mru,a t.!te 7stzSCCzo% -15a- afore-msrtian.d asweiatica+. Both t.he etzengths anG .ibd.iwse of. the study have ixen irdieatedi but rhea these are put inta a bslano., w beleive to lrva 9athered orr of the beat data aets to jLd* the wctense of the purpoc'•.Ad etLtienahip.. we `wrv umble to mtifim an ashciat3cn with the ot oamQSly used sasure of wpoM" ispou.e rnokinJlr but we did firnd an effect doe to eirosars to pas+ive ssokiaq in the housdrold. 'lflse results, altlnu;A only ln partlal a4t+..mint idth tlsse of other .tudies. are mnsistant vith the existenoe of a ssa11 to scderate effect *f passive asoking on lurq Fae+aet risks. Me did oot fi+d an affect ~Iu to aqxiare to passive s'dd+q in the rorkplaos. w fourd. Moevac, a decreasd risk in luaq canosr with its_ssinq aYpoenc+ to pasaive ssioklnq in social clrctstitanoa. Fossible r.asons for sud+ findL+gs have been d.isoassed aCoue. our findinqs swWest tlrt the pr.vtous studies say hsve ouecestiasrad the effect attributaDle to psssive ssdcUq. Suct+ a possiptlity hss been pKpased by thcee .ho oonterd th.c a typial espoeure to passive .addrq is only .quiYSlnK to sseking actively less than orr ei9srett. P.r day (+Rtich in itself omveys risks that are lober than the tw-fold inesase peopo+ed for pwi.e asakicq). 1utAabloqie pcableuis tM.t ry hsve led to sueA ovrsartimt.ian of the effect have bsn+ pointed out in Qwpcer 1. aA -11 as in this chap'.c. As is vell iun.n to epidmuoloqists, the results of one study - or -jen a fe*+ s:udiss-- are seldom suf f inent to anwwr a! ! que:tiou

-.135- :.^q ttae -pIrc:aruvp aetftan a risk factar er; a :eard di.aaase. iie beleive tne present stt"-+/ iss concz:butad to 3 'r.tter :aidsrstardinq of t.he lung carwer risk involved !.n the expasure to passive snakuq. Nevrnleless. re inink re are still far fran Ienwing all the details eeee.+saef to e,et.blish a final judpmmt. FUts e studies will be neoetsaryy to tske that step. Usfi~~~•~z ~~ oz~•~zoz ion ._ ..-0. 82BLIG,7UPlIY 1/ 'fie health cu:seq_uee+oes of smddnq. A rapart of the Surc,+rn irl+eral: 1972. U.S. Dapnc-s+nt of Nealth, E1duc_ ation ard welfare. Public He.1th Servioe. pQ 121-135. 2) S1aF+son. W.J., A preliFinary report on ciqarettt sedcirn ard the l~+cidaice of pre+mturity. Ane::can.Jeurns~ of CYatet:ic3 ard Gyrt q,+loW ;1957:73, 807. - - 3) tae, C. R. . Cfecv of sotlrr's .okiuq habits on birth aal9ht of theis duldren. RO&i. 1 Ibdial jwrz+al; 1959: 2. 673. 61 lait..-itim Reviwi.. S9dcuq. Pcagrrncy. and Derelcpnent of the offspring. Nutritiaf REvie-0s, 31(5):163-115, Mry 1973. S) Rfln.. R. J., Plosey. C.V., Cl.r}oe. G.C.. CRinn, S. Pe_-srtal mtcL,q at hane .d height of dtiilQrs+. aritish Med:cal J«uvl. 1961; 283: 1363. . 6) Ctaaq. C.S.. MyrianthaQaulos, N.C.. !relocs affecting risk of crqenital arlfonations. I Analysis of Fpudemioloqic factncs in oayen_i:al ®slforn.tio[u. Rep;at fran the cvlla5arative Priunetal Projeet. Birth ~efects: Oriqin.l article sezies. 11 (10): 1-19, 21-22, 1975. 7) (meron, P. at. al.. 4t+e health of ssr*ers' ard nanmeloers' dtildren. Journsl of ASlery.. 13(61: 336-341. Ju:e 1969. 81 eonhas. G.S.. Wilson, R. W. Cfiildren"s health in families with ciyare:to asd:ers. Amerimn Jasnsl of Public :tealth. 1981, 71: 290-293. 91 Culleq. J., Respirstocy sy.pLcas in child- and oarental snokinq and pnloigm pcniuction. Bntuh Medical Jourral, 2: 201-201, Aps1l 27, 1974. 10) Marlap, S.. Cavies. A.M., Infaat addssioru to hcepital ard s.tesn+l sffekinq. 1he WuM 1178571s 529-S12, March 30, 1971. -156-

-1S 7- 111 Ncran. I:.J., Rtr: e:fsc3 Of sirnicuw csri:A ~regnaMf. Vr+ Zealand pqedicA, L2cr+a1, Nov 1962, pp S4S--A8. 121 Tager, I.B., Mkisa. S.T., Mwx. A., Roarrl B., " Speixer, F.E. Irnqitudu:al stucy of t:ie e!fects of nnnterral seotux~, on Wimnary fu:c:.iOn in cnildrer.. New i»a1snd Jour:bl Of Mr-'lciae. 309: 699-703, 1.963. 13) Ne::tel, C.1., Budc. C. Effect of 9e*1n9 Ourih9 Pregnancy on the usk of Canotr in Gildran. Jourral of the NatAa±a1 .aaoer 3ns-_V_•ce, 1971; 47: 59-63. 14) Pxatvr-!RS'ti.i. S. , Yu, M. C. , Sentnt, 9.. Nerdarsa:, B. E. V-ni•_=so cz-:powtis and ctildeal tssin tzsorss A osr-oonrrol sy9y. Cacer aesenrchs 1982; 42:.5240-5245. 131 ctuf!eaa61. S.. 1au", B.B., Delaetil E.A., ICin S.Y.l., Delseil, E.S., hllettao J.M. 1QHirv:+r:ta1 factas in the ecioloqy of rhsbdonyoaareom in d'1ldtnod.Jo:unal of the Ns::onsl Cs.-Nors Itut::tt7es 1982. 60: 1C7-113. -- 16) Memirq. M.D.. Carsoll, B.E. SaM epidedoloqic aspecta of leukenaa in children. Ja+r:sl of the N+tionsl Crr9gr Ins'itute. 1957; 19: 1087-1094. 171 Jaffe N., at. al. C1in4_o1 Inv-Atl9atials of the etioinqy of childhmd o>noeis. prsc -w 1lsaoc Cartiar Res, 1978; 191 lel The health aQUSfi>encse of aukinq. 1975. U.S. Daqitsel:t of Health. IIlucMion and Mei[ace. Public Health Serviae. e:e<+ter for Diiaase Crx+trol. 191 Burns, D.M. ScientifiC evide on the hazards of ierrol:ntary seoiup. Joiansl of ftvathi . 38 (3) t 6-7, J:zti 10, 1375. 201 E:tisu:aa. J.E., Godley. E.B. Canotr aorC+lity +nQ:g a reNeser.tativw sasole oi non-seflMess. in the UniteC States durinq 1966-1968. Journal of the litia:nl Ca-xe- L^st.itute. Vol 5„ No 5, 1175=1tH]. NOV :982. 211 Garfinkel, L. Caat wQtality in norssolurs2 ProepecYlve svdy of the harrican Cnxwer Sociec/. Jwrrrl of the Nntion:l Cancar Irrs _ute. Vol 65, No S. 1169-1 7l. Nw 13E0. ~~tzsc(.azOL -139- 22) Rcqot, E., Murssy, J. Carx:er Tortality asnnq nons:okers in an innued qzcup of U.S. Vetersns. JaL-ra1 of the Nstional Cancer Insc:c.ue. Vol 65, No. 5. 1163-lied, Nov 1980. 23) Eiut:-na, J.E.. Risinq 1u:q cancrs eatality aocn3 nuavokesa. Jourrtil of the Nationsl CSnoer Inst.itute, Vol 62(4). f;wil 1979. 211 Carfinkel, L.. Tiw trerde in l:snq conoar ®etaiity arQq rmamnlc.ss ard a r+ots on possive mycinq. J~t ~1 of the Natucnal Car~r DutitUte, Vol 66. No. 6. J4a:e-I981- - 25) Hirayssf, T.. Naeirowainq vives of heavy esoiers have a higher risc of lunq cwj=: A stody from Japan. British Medical .iournai, vol 282, 17 Jan, 1981, pp 183-185. 26) Rarris, J.C.. Du4oud:el. W.L.. Letter to Hirayau. The Bri.ish Medica„l Journal. .283162961= 915, Out J. 1961 (letter). 271 'fridmaulo:, et. al., Lunq eancsr ud passive smokLng. Irtte.-netional Jou-sl of Canoer: 27, 1-4 (1981). 28) 'Yief+apouloe, at. al., Ianq cananr and pwive s+dcinq: Conclusion of Greek study. The Lancet, Sept 17, 1983 pp 677-67e (lecter). 29) 1o1ayo Corrap. at. al., Pesaiw! 900kinq and l:snq canoar. The Lancet, Sept 10, 1983, 59S-S97. 30) (;srfinkel, L. at. al., Itrvolintary aiokinq and lia+q c.ancer: A Cdse-Co<1C2'ol st:Jdy. JOlcra1 of 5~e Na`1olMl Cdl10rS IflstlRlte, Vol 75, No 3, Sept 85, 463-469. 311 fantel, N.. Ifnrolunvry seokinq ard lu+q osnoer. -Sone object lessons. (Ietter). Jasnal of the Na:iotil Cancer Instittte. Vol. 76, No 6, Jwr;-I°86. 321 Akitr, S., Kato, H, Blot, w. Psssiwt ssokinq and 1:nq cwnoer amQ+9 3apenese womrcn. Carcer Resei-1. 46. 4604-4e07, Sept 06. 33) Dalaqer. N.A., et. il. The relation of a asslve s:ocu:c to Cu:q carxMr. Cnc•z 4eha:cl, 46, :eJe-4811. Sept 86.

-160- 341 ua5et. G. C., S+yndar, E. L., hsq canonr in non-artoRers. Canoez. rtsrch 1. 1984, vbl 33; 1214-112l. 351 Saadler, D.P., Wilm¢. A.J., E\rerson, R.B., Clsula:ive ef:ects of life_ume paasive smoking on onozr risk. preliuunary ;.oavuru:.atlon. '1Tie larcrt, ttibruary 9, 1985, 312-314. 3() Sandl.a, D.P., Ever+o+, R.9., vtiloa, A.J., 8zoeder, J.P. Cansc:t risk in adiithocrd fra early life exposure to pn_ restts' tnici+.;. %m:r_on :.aanal of ?mlic Heclth, Msy 1985, 4to1 65, tb 5. 487-492. - - 37) yvdler, D.P., Elrerswn, R.D., Milctac. A.J., Pasaive srokinq in adulthood and auxxs rialc. Merican Josnal of Fbideuuloqy. W1 121. !b. 1. 1985, 37-48. 381 7.S. Department of Naalth and Ba.n Servioes: '8 . health cu+saqueces of sed:irq -ca-oss: a repart of the Slayeon General. OtM Pub. !b. (PR5182-50179, re:ahsqton, D.C.. Govt Prurting Oftioe, 1982. 39) Hoeqq. U.R. Clgarette svlat in Eloeed spa.a. Qwizvrwerctal He.lth Persoecrt.ive. 2, 117-128 (1972). 401 !os R., et. al.. -Exc:reGion of eutaqw in hurn url-ne after PWi`R smak-9. Canoer Lsttery, 19 (1983) 85-90. 41) Hot*J.nn. et. al. '0.ib.ans sidastrea. sfdoa OpEyce by nrn- sdr.rs. P=everrtive tiedicine. 13. 60l-617 (1954). 42) Hinds, W.C. ard PL-at. M.V. Oonoertratiorr of rticotine ard totacoo s+:ohe in paDlic plaees. Hev pgland Journsl of 4mdic'.nr, 292, 844-845, 1975. 451 Jarvis. M.J., Servn thiocyanate in passive sssoicinq. ^.fr [anoet, Jer.w--y 19, 1°85, p!69 llertez). 461 f2iitnraal. Bru_ thinq other pecple's ssdoe. 8ritish Medical Journal. !b 6135, p453, Aug 12, 1985. - 47) Meetinq rr{s~t. EYght itRe_*natirnal MMtinq or+ N'altTCao oocpo:ud,;. Jnsrenca, biolaqia:l effeCts at+d relevancr to tasan cvnces. cancrr Brsftarcn. 44. 1301-1304. !lardf. 1984. 48i U.S. Departtse<+t of Health and llUMn 8enriQs- Pulslie Rea1t'f Servioe. 'L'ie health oQSSecRrencee of stca".. A=eport of Uae 9,i.-yeoi Gencal. 1982. OEffS Puhlimti,an No (R6) 82-50179. 49) Bnarevern R.D.,. 8uffmann. 0. Qhesicsl sbutiiss on tahemo s0loe. Lt:t. Analysis of volatila nitsc.anvrs in tobscm a®ke ard polluted irdoor enrvuvaents. Int wsler. t.A.. Griciute, L., Casteqnar0. F7. , eds. 1'..`7vifCrfM_ftCSl AfpeC'•.s of IMR1tZVeo .. ocapamis. INRC (Scirst::.tic P:;oliotims fb. 19) Lyons: iBD. 1978: 343-356. SO) "Docty, P. M.. Averitt, J.H., and oriffith, R.d. C:arYnt status of the oore ser+ice srakinq beh+viour studies. truversity of Itenturhy '[bdsr.'co and 6e.alth wbrlcshap Confere:cs Aeport . Mo 4. 1973. 10-69. 51) ltiars+ton, R.E. (Bl.) Srdan9 8elrvias. Cnur-hi11-Livingatm. Iardon, 1978. S2) Ridaert. W.S., et. a1. tgtin+tlnq the hasards of 'lesa ha:aidoiu• cigarettes III. A study of the affeR.s of various aioku:q co:diuo:u en yields cf hyiroqen cyanide .nd cigarette tar. J; +IUoc, f-?uLrrn. F1erlth. 12(1). 1983,. 39-. . 43) !f:! fod C., r{awkins, L.B., Aruup, P. lKpoaue of p&.js tve mokers to tolacUo ssrohe cvssritvelKa. Ltt ArcA Oec1 D,icon 53) :,ohnyan. M.R., 4t:e pµro9rJieels physion_c3re*ical na=ue of ) (1971) 1. 7Ubecvo ?ci nd He..lth, 42: 21-19, 1978. . v. :absa o sQtyce. Rec. 44) TTiebiq r., tobe , M,R. Ildpor air pollution bv ssoloe mucituents. A surveY. P:'everttave Medici.-s. l3: i70-581, (1984). 54) Izard, C., et. a1. 2rSlus+ce des psr..etsas de fumaqe wr 1'activite gerv'!=i; + de ls phese qaze.ae de ciquetu aesuree 77 (1980) Rrs Rutat l t , .. euvre. . asnn et La sur 1e 1yRphocytc 341. ~s~zsE~zoz ou....Am.

-161- 551 Andetson. w. ti. 3Lis+.-. n. aud Ave:: ;i.urr`tation of serok.anq in p;:enrs .ntn srocinq relarad eisaases. Chest. t,bt 74 !3), 115. 1976. IAoec.-act). 56) Matsclnrra. S.. :yrwato. T., HaezJasu. R.. et. al, E- fecrs of enviranentai rebaoco sreae or uriuury coc.uiine exxetion in nati+nokers. Eridenoe for passive soomdng. Ner eTrlerd Jouznal of Nedicine. Sept 27, 1984. 8.'8-332. 57) wld. N.J.. eoreMai, ,T.. Frlley, A., et.. al. Urinary wt.inine as .wr/o•r of tre.t:wx~, other (aecple's toh.= suoko. 'ie Lu+eet. Janury 28, 1981, 230-231. S8) Weld, N.. Ritcfiie,. C. Validation of studfss of lunq canoer in *;xr-scokazs saszied to sakria. 7Ae Lrxet. 1ay U, 1984. 1067. S91 Sftiyern. K., Tamin.to. T.. Ritaro, N., et. a1. Effects of divitvnr_mal tohrc=v ssoh on uri>ary oDtinin excretion in non- s®kara. Ne.r Engiand Journal of Medie'ne. Sept 27, 1964. 828-832. 6L1 Jarvis M.. 9lanstail-f+sfoe. 8.. Faysraband. C.. VMsey. C., Salloojee. Y. 6:ochemial earkers of seoiae ahscsption and self- reported exposarre to Qassive sokfnq. Jarrne! of 1ridemiologY and Ccannc.itv Heelth. 1984, 38: 335-339. _ 61) Hiller. F.C.. Me Qwsker. R.T.. Ms+sder. N.K., ttilson. J.D., 0one, R.C. Depositian of sidestrer smke in the lron ;eepiratory tract. Anet Rev Aesp Ois. 125, 406-608, 1982. 62) Riller, F.C.. !t. al. Depositioti of sidsstrsrm ci9Rret•te smdce in the hv.n respiratory tract. Fteventive Medieine. 13: 602-607, 1984. 63) Mitdhell. R.I. Corrsol!ed srastrsent of ssols pa_r_ticles retention in the rrspiratasy tract. Awr Rev pesp Die. 95, 526- 53J 641 Rep.c.. J.L.. tavrey. A.N. lidoor air pollut.ion, tobaom IIrJce, and aubllc health. Scier+oe. 208s 464-472, 13E0. 55) 'Acue. C. pu.ntitatise aspsets of passive ax*.inq md luq Jno•s. Preventive Med.icine. 13: 696-704, 1984. SsVzSCCzoZ -162- 661 Lyrrih. H.3'., et. al. Genetic aid ssokinq ass..-+ated c_anoe_r_s_ . A sudy of 485 f.ulies. Cancer. April 15.. 1986. 67) Dties, A.L. Brondnye:ue prcirrae in dav:ic obetructive piilswry disease. Joianel of the 1Mm_rican Mediml Assaciation. 1976, 235: 621-622. 68) Rruffsem. F.. 'l+eiaier, J-F.. ¢iol. P. Adult pesaive ssokinq e_e aae-low in thr lt=e ersvizv.ant: A risk factor for chrv~.i luutation. Pme:_ican Journal of Ebidenuolooy. Vb1 11'1, No 3. 269-280. 1983. - 69) Nhite, J.R., EYo_ eb. 9.l. Saal l-ti..'ways dysfunction in nm- suo_kers d:rvtically exVoeai to tmecco scdae. New Qglard Jour-.wal of Madicine. 302 (13): 720-723, Murh 27, 1980. - 70) CMan Yue. la Han-1(ian. 4e effects of passive seokiiq on ctu ldren 's pulmon.zy function in Slwngai. JMierican .Jaurra 1 of Puo11c Health. Pey 86, Ybi 76, No S. - 71) Myndrs E-L.. Goodman. N.T. S7eakinq and tM Canrs: Sosse uruesolvnd issuas. EDideiuoleyic Reviers. Vbl 5, 177-207, 1983. 72) Hollard, R.R.. Roslovski. E.J., Booker. L. 7he effects of cigarette ssolae on the r..piratory syst.em of the raChit. A final report. Cadcrs. 16(5):612-615. !hy 1963. 73) SaFfiotti. U. Dq+esinereral raspiratory trac earcinogeneais.. es a in E]pnrimental Tutor Resaartri. 11i 302-333.. 1969. 74) Mynder. c.4., ltn[fs.rn, D. Tobacto and toEaeco seolce. Stuiies in expennent41 ca=ino4er+esis. Ner York AcarieRic Press, 1976, 730 pp. 75) Shuhik. P. '[fie use of the syriar. golden h.tster in chtonic txc?ty testinq. In. Bmt%:rquei, T(Biit_orl. Ptoqress in Deperimntal 'itnor Rexu,--h Patholoqy of the syrian namster. Vol 16, Basel. S. Rargor. 76) GritfitA, R.B., and H.ncaclc, R. Sin:ltan.ous ..ins*resw- sid-e.a sooiu expoase systema. I. 1>>uiptent and iTacedum. 'ItoucvioW, 34: 123-138, 1985. -

-.163 77) Orris. L.. Var. Duuren, B.L.r )osl+k. A.S.. Nelson. N.. Sd_ autt P.L. Cie «ccinugenicity for souse skin and the arontic hyd:vrsrm: emwrt of cigarettrsroke modetiacea. Joaural of L1e Nati=nal Cancer Insti:vte. 21(3)c 557-561. Se6C. 1958. 78) Gcecfbery. R.P..r et. al., 2nddaroe of lvtg ra..r.r by cell type: Apapulation-CssO ats)y in tLw R.spehire arw Vamrrn. Journal of the Na ioral Caxlc .itute. Vbl 71, :b 1. Msrch, 1984. 79) Doll. R., et. al., '1!r .iQnifie.not of c.ll type in re;ation to the aet.ioloqy ef 1is:q canoer. X. J. Cenoer., 1957; 111:43-48. 80) Rreyherq, L., Ristoloqlol l:nq cuxer eypes. A maPholoqiall and Eioloqial caxela-.ian. Acts petnoL. Microhiol. acend. Suppl. 157: . 1-92. 1962. -- - 81) Meiss. W.A.. Ro.erwig, M.. Meiss. W. A.. Risk of ltng cancer acooulin~q to M1istoloqie type ard ciqarette dosaqe. Journal of tne Imerican Medial Association. 1972, Vol .'J: '99-0T-. - e !2) Yensz. R., GelLasn, N. A., aud Fsi+utain. A. R., A csrppcaisal of hiatops:hology of lurq cancer and oorrelation of ci11 type with ant.cedent cigarette asolo.. 1Ner. Rnr. Res. Dis. Vol 107: 790-797, 1973. 83) AuerhrCt, 0.. Gartinkal. L.. Farks, V. R., Rislologic tSpe of lunq Canoe: in r.lat,ior: to andcing h.hits, Saar of df.qrosis ard si'.es of setistaxs. QM. 19751 67: 382 - 387. N) R.G., et. al..'lhc chsnginq RistopKholoqy of luq cancr-r. A seviev of 1682 cases. Cancer. 39 14) : 1647 - 1655, 1977. 85) Stayner. L., and wegr.n. D.. 9m*irm, otrwetion and "is•,.cp9t+loluTj of lung nnce•e A o,se-om2rol study with the use of tne Tlurd Nstion.l Canc_1 S;uveys. ,iouraal of the Neticnal Cancer .nstltvte. 'Ml 70, No 3, M.rd:. 19a3. 86) L-Iterrational HiaiOlaqial Classification of Taraurs. world Mea 1 t.1 Ocr,uu ta t.i on. 1992. 4stzsCCzo% -1b4- 87) Holford. T.R.. Wtute.. C.. Relsey. J,1,, Ulcivariate anayszs for mswned cax-aort[iol studies. Mrs:can Journal o_` padmuolo4y. 108(4): 299-307, 1978. 881 Cornfield. J. Gocdrn, T., 9nith. M.: Quant,l respartse t•.a-:es for mWerinertrally :a.cantrolled variables. eull. I+_t. Statist. ingt.. 38: 97-115. 1961. 891 '-Woer. R.G.. Neldez, J.A.. 1T:e_ GLII4 System Release. Qderd: . Naerivl Algoritlne Gcvup. 1978. - 90) Oo11, R.. Petn, R. Mortality in relltian to aDoking: 20 years observations on sa1e 11ratlsh docs.on. 9r. Med. J.: 216051): 1525-1536, Dec 25,. 1976. 91) Roqut. E. , Murray. J.L..Mmoiung ard crnes of de.th aamg U.S. Veterans: 16 veers of obsarvauon. Public Health Reoat.~. 95(3 ): 21 J-222 .. 1by-Ane r 1980. 92) Tried+ar. G.D., Petitti, D.s.. Mwl. R.D.. Ptsvalenoe and carrelates of passive s®kuy. h:erican JotaTwi of Public Health. Apsil, 1983. Vb173, No 4. 93) lJnited Nstions oe®qraphie Yea;N=k. 1978.

l : l 1 • l 1111 1 'i! L Ij 00 ; ~ ~ v © Q O. K .iwd4 4 ® @ E QO~JOE)

H I E• ~ .. s 1 ~ .. s 1" .a ~" l' ,, ~ 1 s' 1' - s= l ir 1 sE 1 .© is 1 i_ 1 - s l" ~= '~ --- Il'~ 1= 1'c l: 1= " lc l" ls 1' 1= 1" ;1s " ls l~ l: . ~F l l 1 1 l 1 1 1 E~ a ~ 1 { ~] . I I - l~ ~ii z 1R 1 ~: 1 ~, l 1t l ~i l 1R l 1' l la l 1- ~ ' 1? l _ ~~ ~s \ j~~ 1 1 1 1 l 1 1 1 1 l 1 .i = ~ . ~ , . l E E 1 je!ee ~ A d 3 i' s' : = I

.. . !B _ t~i1 ~ _. S 1 . L l . f l~ . i 1 . t ~ 1 . : 1 . i i l- . = .. - 1` . ; - 1 . = . r - 1 . 0 G) 0 0 0 0 0 'O 0 O O 3 l' l: l: l: b lb . i~ ~ ' . , . . . . .- .- .- . . . - --w: ~=~-- . l 1 l c 1 l 1- l • 1 1 l . 1 1 1 1 1 l l l 1 . 1= 1 l 1= 1 ' 1- l 1 . 1= l - l 1 ~ l 3 l 1 l .. 1 l 1 ~ l l l. 1 , _ j : = : = _ j - w - : - - R - 0 ~3~ . ~ . . . . . . . ~- . ~ •. . l~i E f ! 75~~~. ~e . p ~ . t le . p 1= ._ f lR .w • 1= .- 0 ~ . . •~ .. • ~ ._ ' lc -.w ~ ' 1•. ,. ~ j.l_ ; ~~ s l, 1. l_ 1. 1=_ lz , l. 1R ~ 1E ~ l: 1. .~-- ' !y~ 1c l: ls 1; l; l; 1= ~ ~ li 1; 1_ 1311 ~ e j ~ l . l ~ l~ ~. l~ l= ~. l~ '1C 1 . 1 1 1 l ;~ !i 1 7 i i ~ = ~ E ~ l' ~' .. o 1'- . 1= . 1'- . 1? . ' 1 ._ 1= .w l. . 1= . w I 1, . '~: . . e e e o , o , o e e e e ti;l- ~ ® ' e ` o = e f ® e ` O e e e e o a~~~l~~~i l 1 - ~ 1 - l= ,l- l. lx 1! . 1? i l~ ls - F F ~ .s

11 ~ `t0 1 . '~. j aaasa ~ 1 1 i 4 e a e e e *1 0 1) e e e 5 e ~ s ~ E ® ® 0 e © e © 0 1? 0 ©

1 8 3 J ~' $ 1 1 . 1 ® ! . I i ~ ~ ~• ~- lllll111 j~tt o oo1ooooo ... i. ~:14 4.Sw'w4 DDDDDIDDD I I d' i .t .i .i : d .: oDDDDDDD d i d i i.: d d DDDDDC]DD .i s.i x s.i s x

. a llll 1 111 1 1111111 l .... . ... . ....... . 9 a C O Q 9 O 1: u i i •: j i 9 C C 01 ® 4 9 l~ 1 x l,a 1 O © ~~ 1 ~ 3.iRt r ~ ~~ attt' ~~ t{ ~, r~ t#£aa ~ j ~ ~~~ ~ ~~ttt t I ~ ~~t : i f;~;' ~ E to E~ a , c ~I ~i p s==3's a~~ ~.