Tobacco Documents Online

LI=TTF_.RS TO THE EDITOR AN ESTIMATE OF ADULT MORTALITY IN THE UNITED STATES FROM PASSIVE SMOKING Dear Editor: P, epace and Lowrey (lgg0a) and Wells (I990) are critical of the points I raised (Lee I990) regarding the paper of Wells (1988). Some further comment is necessary. Repace and Lowrey claim that I have argued the diseases of smoking are not even plausible in non- smokers. This claim is false. As I have clearly stated (Lee 1989a). some effect of enviror.m'~ntal tobacco smoke (ETS) is of course plausible. My view is, and always has been, that it is implausible that any ¢ffcct of ETS might be anything like as large as that indi. cared by simplistic interpretation of the epidemioi- ogy, One reason for my view is th~ massive discr".pancy between the "epidemiological" estimate of risk of lung cancer from ETS (i.e., that based on taking the fragile epldemiological evidence at face value) and the "dosimetric" estimate (i.e., that based on ex- trapolating from the risk in relation to ac tire smoking using estimates of relative exposure of passive and active smokers to relevant smoke constituentS). Put- ting aside issues relating to a possible threshold, the size of this discrepancy is well illustrated when one considers that the epidemiological evidence suggests the excess lung cancer risk in relation to ET$ posure is I0-20% of that from active smoking 1988; 1989b) whereas th.-- dosimetri¢ evidence (Lee 1989a) suggests the factor is more like 0.5% (based on inhaled particulate matt:r or salivary cotinin¢) or 0.05% (based on retained particulate matter). Remarkably, Repace and Lowrcy (1990b) have tried to claim this discrepancy does not exist, noting markable agreement" between risk assessments, whether epidemiologically or dosimetrically based. This totally erroneous conclusion has been reached by an analysis which (1) rejects any low dosimetri- cally based estimate as "being inconsistent with the epidcmiology," (2) accepts a spurious dosimctrically based estimate of Fong (1982) which uses an trcmely high estimate of exposure in passive smok- ers, and O) includes an est-~mate from the NRC report 0150-4|20/91 ~t.r~ : .00 (I986) as if it were dosimetric when in fact it is epidemiological, while ignoring a much lower dosi- metric estimate. Unlike Repace and Lowrey, this discrepancy is admitted by Wells, though his estimate of the mag- nitude of it is less than mine. Wells seeks to explain it in terms of differences in chemistry and physics between direct and passive smoking, though his ex- planation is not persuasive. He notes that particulate matter retention in the lung from ETS exposure tends to be mainly in the alveolar region of the lung, whereas that from direct sin?king is predominantly in the traeheo-bronchial region. Ho@ever, his cal- culation shows that alveolar retention from ETS ex- posure is still only 1% of that from direct smoking, far too low to explain the epidemiology even if alI the lung cancer arose from deposition in the alveoli, which is of course not the case. It also seems diffienh to align Wells' theories with studies by Garfinkel at al. (1985), Pershagen et al. (1987), and Triehopoulos et al. (1983) that suggest an association with squa- mous and small.cell carcinoma rather than with ade- nocarcinoma and large-cell carcinoma, as would be predicted. Wells' argument concerning vapour.phase and tar and nicotine are confusing and not clearly thought out. In the first place, tar, by definition, is the par- tieulate phase. I suppose his hypothesis is that the transfer of some substances from the particulate to the vapour phase in ETS gives rise to a greater than expected risk of disease at noncontact cancer sites and of heart disease. However, if this hypothesis were true, and certain "harmful" elements in the p/~r- ticulate phase in mainstream smoke transfer to the vapour phase in ETS and hence do not impact on the lung, one would then clearly expect that the risk of lung cancer associated with ETS exposure would be even smaller from comparison with smokers' expo- sure than expected from a more simplistic dosimetric comparison, an.d not larger. Also, chemical consider- ations indicate that the majority of compounds likely to transfer from the particulate to the vapour phase would be simple hydrocarbons, with chemical classes such as polyaromatic hydrocarbons and nonvolatile BATCo document for Mayo Clinic 27 March 02

Le~tcrs to the edhor nitrosamines staying in the particulate phase. What class of compounds does Wells have in mind that would be nssooiated with disease at noncontact sites, and volatile enough to transfer from th~ particulate to ~h~ vapour phas~ in ETS? Wells su~s~s that smoking might pro~cct aR~ins~ ~ff¢c~ of ETS. All ~ings ~c pos~ibl~, no doubt, but th~ cvid~nc~ cit~ is sc~cely compelling. The work of ~mmcr ~987), for example, is very unconvinc= inR. ~cdv~on of ~nzymes is no~mally consldcr~d to indic~ ~n i~cr~ased Hs~ of c~nccr, and not the W~lls attempts to reply to my criticisms of his theories concerning susceptible groups by udng ~c s~onsous dcfinltion proposed by Khoary st al. (1989), in which susc~ptibiliw is squared to th~ proportion gotfing a disease by a given ag~ ~n thc absence of competing risks. Susccptibl~ moans capabls of g¢t- ring discaso. By th¢ erroneous definition, pcopt~ who survive Russian roulctt~ would not bs susccpdbl~ to bullets[ Publication bias may not b~ a major issu~ for lung cancer, although there is som~ ~ndircct cvidcnco of its cxist~nc~ (studies in the top third of ~ r~lativc risk distribution are based on significantly lower numbers of cases than sludi~s in th~ bottom third), but I b~l~v~ it could bo oxtromsly important for h~art d~s~as~. Hcr~, thcro is a known large data sol Th~ American Cancer Society's million-p~rson study has published on ETS and lung cancer but has not pub- fished on ETS and h~art dhcas¢ (or any oth~r caus~ of d~ath) d~spi~s having information on many thou- sands of d~aths among married n~v~rsmokcrs. Sinc~ there is a strong likelihood that no association was found (otherwise it would have boon publlshcd) and sinc~ th~r~ arc mor~ heart dis~as~ deaths in that study than in all th~ studi~s W~lls cons~dsrs put togothsr, sovoro publication b~as seems very [~ksly to hav~ occurred. Tho quostion of posslbls confounding in hcar~ dis- ~as~ stud~cs is still an open one. Th~ only ETS studies that hav~ dam on th~ ma~n confounders such as blood prossur~, cholesterol, and body mass, have b~n based on ~xtrcm~ly small numbers of dca~s, probably too low for any sort ot roHabl~ adjustment. The possibility that ETS might cause cancers not associated with direct smoking still s~ams m~n~mal to ms, d~sp~m W~lls' arguments. Th~ dlff~r~ncc ET$ exposure between a smoker ~d a nonsmoker (du~ mainly to th~ smokor's exposure to his own cigarettes) s~ms certain to b~ higher than that twccn a smoker ma~icd to a smoker and a smoker marri~d to a nonsmoker, so any risk associated only with spousal smoking should be viewed with suspi- cion. In fact, accumulated evidence to ~te does not indicate ETS has been consistently associated with any such cancer, chance being a likely explanation for a number of the isolated reports of associations. The question of the extent of bias duc to misclas- slfication of active smoking status is a complex which cannot be adequately dealt with in a letter. Like Wells, I am publishing separately on this sub- ject, using new material, Some comments are, how- ever, relevant. First, I must agree with Wells that some confusion had arisen in my curly work from failure to distinguish th~ proportion of nonsmokers who are truo current smokers from the proportion of ncvcrsmokcrs who are ~u¢ cu~cnt smokers. Since the cpidcmiology concerns nsvcrsmokcrs, misclas- sificafion of current as oxsmokcrs is not relevant. A~ the time when all the early work w~s publishod, by myself (1987), by Wald st al. (1986), or by W¢lls (1986), lhors was in fact no evidence available on misclassificadon of current smokers as ncversmoksrs, the cotinin¢ studies consld¢rcd all r~lating cotininc to cutout smoking habits. (It is strange that includes my cotinin¢ study in his evidence as no questions w¢rc ask¢d on past smoking habits.) It is only recently that sores more strictly relevant data have come to light, but th¢ results arc r~thor variable. Using this and othor data on the proportion of smok¢rs denying ¢vcr having smoked, I ¢stimat~ that misclassification could explain the whole of ~hc as- sociation soon with spousal smoking in U.S. men and a substantial part (soma 60%) of the association s~en in U.S. woman. Misclassification at the level ob- served would also cause a substantial bias for Asian men but would not caus~ material bias for Asian wom~n. How¢vor, ~videncc on m]sclassificafion ra~cs in Asian women is sparse, though there ars some indications it may b~ very high, e.g., the study of Saho (1988) which found that 55% of Tokyo Univcr- slty student women who smoked rcport¢d their fami- ly knew nothing about it. One problem in the whole issue that has not been fully rccogniscd is the dcpcnd- ¢nc¢ of the misclassification rate on th¢ particular circumstances of the study, and the nccd to correct each study individually for its own estimated mis- classification rate. Correcting overall (morn-analy- sis} relative risk estimates for overall estimates of misclassification can lead to totally fallacious suits. Finally, I thank Dr. Hirayama for his comments on ~¢ discrepancy b~twe~n his heart d[scasc reports 1981 and 1984. It is important m note that, else- where, Hizayama (1990} has now noted that h~s 19el BATCo document for Mayo Clinic 27 March 02

L~,r~re tn the editor data were in error and has published revised figures which are consistent with his 1984 data. Peter N. Lee P.N. Lee Statistics and Computing Ltd. Surrey, United Kingdom REFERENCES Fang, P. Th~ hazard of eisamuc smoke to nonsmokers. L Biol. Phys. 10:65-73; 1952. Garflnk¢l, L: Auerbach. O.; louben, L. ]nvolunta~ s~oklng and lung cancer: a case control stud~. J. ~ac. ~aacer Inst. 75:4~3- 369; 198~. measurement Of ;usceptlbilhy in cpld~mio[ogi¢ studi~s. Am. EpidcmioL 129:193-I9~ 1989. Lee. P.~. Passive smoking and lung cancer association: a result bias? Human Toxicol. 6:5[7-524; [987. Lee, E~. Miscllssificsdon of smoking habits and Furtive smok. lug. A ~evicw of the e~dencc. [recreational Archives of cupstion~l Hethh. Berlin: Sp~ng¢r-Verlag; Lee, P.N. Problems in inte~rcCing epldemiological data. In: Mohr, U., cd-in-c~ef, Assessmcm of inhatacion hazaMt. Ber- lin: Sp~nger-Verisg; I~89a:49-~9. ~¢. EN. Passive smoking ted lung e~ncer. Fret Or fiction? In: B~cVs. C.J.; Counois. Y4 Geysers, M., eds. Present and future of indoor sir quslhy. Amsterdam: Elsevl¢~ 1989b:119-128. Lee. P..~. An estimate of adult monalily in th= United States from passive smoking: a response. Environ. Int. 16:179-1St; 1990. NRC ~on*l R¢lear~h Counc~). Env]tonmcn~*l Iobacco smoke, messing ¢~po~utez and a~sessing bcal~h affects. Washing- tea, DC: ~t~ioad Academy Prc~s; 1986. Percheron, O.; H~bec, ~; Svcassoa, C. P~s~ive ~moklng and lung can*erie Swedish women. Am. J- ~pidcmioL 125:17.24; 1987. Romper, H. Pssslvcly inMl~d to~cco smoke: a chs~cnge to tautology and preventive medic~e. Arch. To,coL 61 19S7. Repine. LL.; Lowrcy, A.H. Rebuutl to Lec~ttzcnstc~ ts~ on pusive smogs ~k. Environ. In~ 16:182-1~3; 1990s. Rcpsce, LL4 Lowtey, A.H. ~xk assessment methodologies for passive smoking-{nduccd lung canc~c Risk Anatysls t0:27-37; lgg0b. S~to, ~. Smo~ng among ~oung women in ~tpzn. In: Aoki, M.; Histmichl, S.; Tomintga, S.. cds. Smoking and Health 1987. Amsterdam: £~¢c~ta Medics: tgxg:St7-529. Tfic~opoalos. D.; Kalandidi, A.; Sparros, L. Lung cancer sad passive smoking. Conclusion of ~¢ Greek study. Lancet (ii):677- 678; Wild, ~.J.; Nanchahal. K.; ~ompzoa, S.G.: Cuckle. II.$. Does br:s~in8 o~cr people's tobacco smoke cause lung cenccr~ Med. L 293:1217-1222; 1986. Wells, A.L Misdasslflcttion t~ t hctor in passive smoking risk. Lancet (Iii):638; I9g6. Wells, A.J. An ettimtte of tduh mortality in the United States from pal,ire smoking. Environ. Int. 14:249-265; 1988a. Wells, A.J. An estimate of tdu[t reenact7 in ~e Unltcd States from ptstlve tmoklng: a r~spoase to c~tici~m. Environ. Int. 16:1~7.192; 1990, PASSIVE SMOKING: A REPW Dear l~ditor: This letter is in reference to Lee's criticism flee 199I) of my paper (Khoury 1989}. I believe the def- inition given in that paper is not erroneous if the faftor of interest and the exposure dose are adequate. ly defined. For example, for a given level of smoking for a certain humber of years, x% of people will develop a specific disease from a sufficient cause Involving ~moking. As discussed in the paper, this measure is affected by many variables inulu~iag dose. duration, type of cigarettes, et~. With respect to ~he example given by Le~, on who is susceptible to Russian mulette, the i~ea of not being sasceptibl~ to bullets if you survivv Russian roulette is obviously mixing two kinds of exposure and different dos~s. Certainly, given enough plays of Russian roulette, everyone would succumb. There- fore, everyone is susceptible to bullets from Russian roulette after numerous trials. One limitation of the sufficient cause model is the deterministic nature of the model. Nevertheless, I believe that the estimation of the proportion of snsceptibles is a useful adjunct to other measures of association derived in epidemi- ologieal studies. Muin J. Khoury Birth Defects and Genetic Diseases Branch Center for Environmental Health and Injury Control Atlanta. GA 30333 REFERENCES Khoery. M.I.; Flaoders. W,D.: Greenland. S.; Adams, MJ. On the measurement of susceptibility in epldemlologic studies. Am. L Epidemiol. 129:183-190; 1989. t.c¢, P.N. An estimate of adult monallty in the United States from passive ~tokiug (letter). Environ. lut. 17:37g-~gt; t99t. 3ATCo document for Mayo Clinic 27 March 02

data were in error and has published revised figures which are consistent with his 1984 data. Peter N. Lee P.N. Lee Statistics and Computing Ltd. Surrey. United Kingdom REFERENOES Funs, P. The hazard of cigarette smoke to nonsmokers. J'. Biol. Phys. 10:6~-73; 1982. Os~nket, L.; Au~rbacb. 04 ~ouben. L. ~nvotumz~ lun~ c~nc¢~ ~ ~se comrol s~udy. L N~. C~nc=r ~a~. 75:463- 369; 1985. Hir~y~, T, P~s~ve smoking. N.Z, Mcd, L 10~:~4~ 1990. m~a~urement o[ ~asccptJbili~y in cpJdcmiol~c ~mdie~. Am. E#demioL 129:183-19~ t989. ~e~, E~. ~ss~vc smok~n~ ~n~ ~ng c~ce~ ess~c~at~on: b£t;? Human ToxfcoL 6:~ 7-~Z~; [997. ~e, P.N. ~isc]~s~ific~fion of smoking b~bh~ and pas~iw ing. A review of the evidence. InC=mational &rchives of Oc- cupational ~eahh. Berlin: SpOnger-Voting; l~8g. ~ohr, U., ed.in.cMcf, Asscssmcn: Of ~hal~tion h~zards. li~: SpOnger-Voting; Lee. P.N. Pa~vc smoking and lung cancen Fact or ficdon~ In: ~{cv~, C.~.; Cou~ois~ ~; GOVaC~, ~., ed~, ~t~scnt an~ future o[ indoor air quality, A~zerdam: Elsevier; l ~8~b:] 19-128. Lee, P.N. An nstimatn of ~dult morta:hy in tho United States from passiw, smoking; a response. Environ. Int. 16:179-181; 1~90. • NRC (~fion~l Re~arch Co~¢i]). EnvJronmemal tobacco ~moke. messudng exposures snd a~c~sin~ ~eal~ effectt. W*shing- Fer~agen, G.; ~mb¢c,~; Sven~on, C. ~assiv~ smoking and ~ung c~nccrin Swc~sh womb. Am. ~. EpldemloL 125:17-24; ]987. ~emmcr. H. Passively ~hsled tobacco smok~: lo~co~ogy and preventive mcdlclnc. Arch. To.coL 61:89-104; 198T, Rcp*~, LL.; Lowrcy, A.H. Rcbund Io L~c~slz~ns~ein common. t*~ on passlvc smok~& risk. Environ. ~t. 16:1~2-183; Rcpscc, L~; ~wrcy, A.H, ~sk assessment methodologies for passive smok~ng-~ducc~ lung cancer. R~sk 1990b. Saho, ~. Smoking smon~ young wom~n in Hisamiohi. $4 Tom.ass, S., cds. Smoking sad Hcal~ 1987. Amsterdam: Excepts ~edica; T~=hopoulos, D.; Kalandld], A.; $ps::o~, passive smoking. Conclasion of Ihc Ot:~k study. Lancet (ii}:677- 6~8; brc~zhin~ oxhcr people's wbacco s~okc cause long cancer? Br. Mcd. L 293:1217-1222: 1986. Wclls, A.L Mi~clsssificsdoa a~ s fzc~ar in passive smoking ~sk. Wells. A.L An esfima~ of *duh mor:ality from passive smoking. Environ. ]~. 14:249-26~; 1988a. Wells. ~.L An c~cimete of ~duk mo~alhy hum p,s£vc ~mokiaa: • response ~o c~£¢~sm. Environ. Inc. PASSIVE SMOKING: A REPLY Dear This letter is in ret'erence to Lee's criticism (Lee 1991) of my paper (Khoury 1989). I believe the def- inifion given in gloat paper is not erroneous if factor of interest and th~ exposure dose are adequate- ly d~fined. For ~xample. for a given level of smoking for a c,train number of years, x% of people will develop a sp,cific disease from a sufficient caus~ involving smoking. As discuss,d in th~ papeL this measure is affected by many variables including dos~, duration, gyp, of cigarzttes, etc. With respect to th~ ,xamplo given b~ L*o, on who is susc,ptible to Russian roulette, th~ idea of not b~ing susceptibl~ to bullets if you survivc Russian roulctt~ i, obviously mixing two kinds of exposure and different doses. Certainly. glv~n ~nough plaFs of Russian foul*tee, everyone would succumb. fore, ¢veryon, is susceptible to buIle~ from Russian roulette after numerous trials. One limitation of the sufficient cause model is the dzterministic nature of the model. Nevertheless, I believe that th~ estimation of the proportion of susc~ptiblts is a useful adjunct to othcr measur,s of association derived in ~pidemi- ological studies. Muin ]. Khoury Birth Defects and G~n=tic Diseases Branch C~nter for Environmental Health and Injury Control Atlanta, GA 30333 REFERENCES Khout~. M,L: Handers. W.D.: Greer.!sad, S.; Adams. M.L On the measurement of sutceptlbility in cpidemiologio studies. Am. I. Epidcmiol. 129:1~-190; 1989. [.ce. P.N. An estimate of adult mortaXty in ~e United Stat~t from passive smoking ~:tt¢O. Environ. [at. 17:379-381; 1991. BATCo document for Mayo Clinic 27 March 02

PASSIVE SMOKING AND HEART DISEASE Dear Editor: Olamz (1990) criticized Lee at al. (1985) for draw- ing negative conclusions about the association tween passive smoking and heart disease based on a study which he, Glantz, stated to have only 3% power (to detect a relative risk of 1.2)o the ]ow¢s~ o~ zny such study. In roply, L~o (19~0) po~t~d out ~a= his conclusions were bas¢d on the overall evidence, hoe just one study, and ~haL in any case, his study h~ power of 12%. higher {hzn m~y o£ th¢ ocher studies. This conclusion contrasts with tha~ o£ Glantz and Parml¢y (]991), who ~nd =hat passive smoking ni~ican¢ly increases ~hc ris~ o~ hcarl disease in smokers. To resolve the discrepancy between the power values, we have r¢vlcwed ~ach other's lion procedures. ~hil¢ th¢~¢ are statis~c~ ties in power calculations arisin~ ~rom ~h¢ limRed and dff~¢~ng ma~¢rizl zvzilabl~ in the v~ious s[udics, two points clearly emerged. First, L~e had based his calculations on the total data £or Ih¢ sc~es combined, whereas Ghmz considered daea for £cmalcs and mal~s separately. Second, there was an un~oreuna¢¢ e~or the computer program used by Glantz which h~s original power calculations inco~cct. Gtantz sub- sequcntly corrected his program and we now agree tha~ th~ power o~ toe's study (around 4-~% ~or males or females individually and 10-12% for the combined sexes) is quite low, and substantially loss than that of the Hirayama and Helsing studios. It is, however, greater than many of the other studies and is not the least powerful study of passive smoking and heart disease, Stanton A. Glantz, Ph.D. Professor of Medicine University of California San Francisco. CA 94143 and Peter N. Lee PN Leo Statistics and Computing Surrey, England REFERENCES Giantz. S. Risk =ssessmen.~, of pa=siv¢ =rnoklns (letter). Eaviror~. In. 17:~9; ¢Fidtmiology, phy=iology, and biochemlst~. Circulation g3:1- 12; 1991. 1990. dis*as*s, Br, J. Canc*r 54:97-105; 1986. AN ESTIMATE OF AOULT MORTALITY IN THE UNITED STATES FROM PASSIVE SMOKING; A RESPONSE TO CRITICISM Dear Editor: Lee (1991) has made further comment on my paper (Wells 1988) in Environment International. It is hear- toning that Lee now agrees that his earlier analysts (Lee 1997, 1988) of the bias introduced by misclas- sification of smokers as nevcrsmokers were not carefully done. He agrees that he had confused the proportion of nonsmokers (neversmokers plus smokers) who are true current smokers with the proportion of neversmokers who are true current smokers. This "confusion* approximately doubles the calculated bias. Also doubling the bias is Lee's prac- tice, as noted curlier (Wells I990a), of mixing male data based on nevorsmokers with female data based on neversmokers to calculate the bias for studies of females. Le, also now agrees that he should have calculated a bias for each study separately instead of applying U.K. smokers' risk and smoker prevalence to studies from other areas of the world where these values are much lower, a practice that again substan- tially inflated his calculated bias. Lee has retreated from his earlier position that misclassification of smokers as nonsmokers is a major source of bias to a position that it does not cause material bias for Asian women but still causes 60,% of the association observed in U.S. women. Calculations that Professor Waiter Stewart of lohn Hopkins University and I have carried out (Wells 1990b} indicate that misclas- sification of smokers as neversmokers may account for up to 2,% of the association observed in Asian women and up to 28,% of the association observed in U.S. women. For all 28 studies worldwide, that are now available on funnies, this type of bias may ac- count for up to 9% of the association, hardly a "major" source. There are lWO main reasons why Leo gels 60,% and we get 28% for U.S. women. Both we and Lee rely on general community surveys of two types. In one, people are asked if they smoke. Then shortly there. after, they are tested for cotinine level. In the other. 0 "',4 BATCo document for Mayo Clinic 27 March 02

Letters to ~h~ editor 353 peopl~ arc asked ~t two different points in time if they had eve~ smoked, and discordant answers are noted. However, l) Lee continues to mix mate data with female data from th~s~ studies, and 2) hc suspect data sources. For ~xampl~, h~ (Lee 1989) r~lies on the larg~ Haddo~ (1987) s~ud~ whereas authors of thai paper strongly recommend that the eofinlne data in i~ no~ be used for miselassifi¢a~ioa estimates re passive smoking (private communlca- ~on from Dr. Knight). Instead, they recommend the data in their 1986 aug 1988 papers (Hagdow 1986; Haddow 1988) where th~ data gathering was mu~h closer to wha~ would ~cur in epidemiological s~udies on passive smoking. So far, Le~ has ignored the 1988 paper (Lee 1989). Lee also implies that misclas~ifica- tion rates among Asian women are higher than among other women, but Dr. Koo (~oo 1990), who par- ticipated in and had access to the cotinine dam from th~ rec~nt IARC study in 13 ~eas around the world. misclassificatlon rat~s and such rat~s elsewhere. There is also rcason to baliev¢ that ewn th~ bias estlmatcs tha~ w~ hav~ ma~¢ arc too high. ~or ~x.amplc (W~Ils 1990b). th~ genera[ community surveys that we hav~ used indicate that 5.7% of eversmok~rs arc exsmokers who ~ misclassified as newrsmokcrs, whereas in th~ five passive-smok- ing cpidemiological s~udics whets misclassification estimates were made by the authors, only about 1.0% of ~vetsmokers are exsmok~rs found to be misclas- sifted. The o~her evidence, eha~ the misclassification factors w~ have used are probabIy too high, comes from the studies on mal~s. In ¢h¢ usual male cohort, there ate fat fewer ~pott~d n~vctsmok¢rs (17~ 35%) ~haa in fcmal¢ cohorts (41% to 86%). In calculating the passive risk corrected for smok~r- misclassification bias, it is n~cessary to sub~ract th, current and ~xsmokets misclassificd as smok¢cs f~om the number of repotted n~versmo- kers. Whca one uses the misclassificaHon factors gonera~cd from ~hc communiW survc~s, ~h¢ smoktr co~ec¢ion dcl~doas ~xc~¢d ~ total n~ber of reported n~v~rsmokers for three of the el,yen male cohorts and drive ~c corr¢cted risk substantially below unity for another four. This can only m~an that th¢ misclassificadon factors derived from th~ com- munhy surveys arc ¢oo high. Thus. until we can better misclassificadon data on which to bus, mal~ co=¢ction calculations, it is probably b~s¢ to do as the U.S. Environmental Protection Agency h~ don, ~SEPA 1990). namely to us~ the corrected femal~ passive smoking risk as a proxy for the mal~ cor- x~ct¢d risk. ~{caawhiIe. Let's calculations of the p~ of th~ obserwd passive smoking association due to smoker miscla'ssificatlon for either U.S. men (60%) or Asian men (substantial) can b~ regarded as meaning- less. As Lee say.s, there is a discrepancy between the relative risks one obsorves in the epidemiology for various diseases associated with passive smoking and th~ relative risks one calculates by ratioing down dosimctrically from the risks associatod with ac- tlw smoking. Bu~ is this a r~ason ~o throw out cpidcmioIogy in favor of a calculated effect? I think found that can explain the dlffcrcnccs. A b~ttcr as- sumption is that the dos~ rcspons~ curves for passiv~ smoking and ac~ivc smoking arc different du~ to thr~ main reasons, I) th~ natur~ of ~h~ smoke d~posi- don is different. 2} ther~ is a balance of toxic and pro~cctive effects thai varies by disease and is dif- ferent for active and passiv~ smoking, and 3) there a diftcr~ncc in awrag~ susc~pdbilhy for each dis- case between the r~ladvcly small proportion of the population who di~ of passive smoking and relatively much larger proportion who die of ac~iv~ smoking. GIantz and Parmley (I991) have published a com- prehensive review of the effects [ha~ passive smoking has on the functions of the hearz and circulatory sTs~ms. They demons~ate conclusively tha~ on~ can- not predict th~ effects of paszive smoking on hcat~ disease by using calcuIations based on dose relative to active smoking. For cancers other ~han lung th~r~ at, not onIy the overall studies referenced ~arl~er (Wells 1988. bur a number o£ szudi~s of specific cancer sites. Thcr~ ar~ now sewn studies (Buckley I981; Brown 1982; Hcllbcrg 1983; Sandier 198~; Hirayama, priva~ communication; Slaztcr~ 1989; Buffer 1990 and privat~ communication) of passiv~ smoking and cancer of the cervix with 456 cas~s of ncwrsmokcrs and r~Iadv, r~ ranging from 1.25 ro 4.1 [or an average otabout 1.5. Th~ studies on breast cancer and passive smok- ing hav~ recently been summ~ized ~llS 1991) show- ing a r~lativc risk of 1A. In addition, studies exist for several o~h~r sites, and my calculations of total deaths by si~ agree remarkably well with my es~imates from ~hc total cancer data. The cffec~ ot vapor-phas~ ~ar in ET~ may hav~ a bearing on th~ rcIativcly high xlsks obsctwd for non-lung or non.contact sit~s. Le, notwithstanding, t~ is usually d~fin~d ~ thatparz of"freshly gcn~tcd" cigarctt~ smoke, less water and nicotine, that r~incd on a Cambridge filt~r. What Pritchard et a[. (1988) found was tha~ 70% of sid~stream t~, so E ATCo document for Mayo Clinic 27 March 02

Lcucrs to i~c defined, evaporates into the air as ETS ages. Their marker, cetyl iodide, has a boiling point of 3g0°C, placing it between the three-ring and four-ring poiyaromatic hydrocarbons. The particulate phase has a low retention factor in the respiratory tract. about 15% (Wells 1958), but th~ tar constituents in ~e vapor phase, li~ o~hac organic vapors (Bond th~ r~spiratory tract. Thus there is the probability tha~ the to~al r~t~ntion og vapor could be ~70/30) (100/l~) = most a~ ~h~ l.ng car~inog~n~city, in t~ compounds og four or more rings, although about one sixth is in th~ two-and throb-ring fraction. Their cut poinc is som~whaC lower than Pritchard's so we might guess that about one-fourth of th~ lung carcinogenicity is coming from ~ho vapor phas~ og ~TS. Th~ higher molecular wclght carcinogens, those over 200 and contained in the HTS particles, ar~ retaincd got long periods in ~h~ lung ~Hcnderson et M. 1988~, ~hus contributing carcinogcnicity. However, the vapor pha~ material with molecular weights og I00 to 200, and including such compounds as quinolinc, m~thyl quinolinc, b~nzoquinoline, phenan~hriden~, nornicocinc, ~-naphthyl amine, nitroso pyrolldine, ni~oso nomicotine, pyr~ne, fluoran- rhone, phenol, ~h~ cresols. 2,4-dime~yl phenol, cate- chol and th~ methyl ca/cchols som~ ~ype of carcinogenic activity) would clear groin th~ lung into the body fluids mor~ quickly and would add to the potential got cancers al other sites. What we don't know, but suspect, is that much o~ the ¢aroinogenicity og mainstream smok~ is lost b~causo so many o~ the larg~ mains~eam p~icles ar~ cleared rapidly into th~ mouth and swallow~d, only ~o ~o out with th~ g~ces, unme~bolized and unabsorbed anywhere in the body (see, for ~xamplc, Bond et al. (1986) on loss og inhaled nitropyrene particl~s to ~cc~s). In any cas~, ~hc presence o~ a large amount og vapor-phase {ar en~i~ies accompanying the particles og HTS enhances ic~ lung carclnogenicity and in- croascs~to an ~v~n greater cx~cnt--th~ poten~al for cancer elsewhere in ~e body. believe that there is a rang~ og susceptibility among individuals {o smoking-induced cancer in ~ ligh~ of all lh~ recent work on tumor suppressor Kcnes and related genetic research. Dr. Khoury (Khou~ 1991~ has r~pli~d s~paratcly to criticism of his ~ork on individual susceptibility. Howewr, h smn~ to reason that if th~r~ is a rang~ of susceptibility to cancer from cigare~ smoke--and all th~ recent work indicates ~hat the:= i~thc most susccptiblc individuals are going to die younger and/or will be affected by luwer doses than a larger but, on average, less susceptible group. We may not know exactly how to quantify this difference yet, but intuitively such a difference must exist. Re publication bias, Lee continues to hope that the American Cancer Society will publish its data on ETS and heart disease. P~rhaps ~h~ reason {hey haven't, is besl d~scribed in Dr. Cuyler Hammond's words since h~ was responsible for gathering ~hos~ dam in {h~ firs~ place. He says (Hammond 19SI) ~h~ h~ "would hav~ liked ~o eslima~c lung cancer dcz~h tat~ in rcla[ion m amoun{of"passiv~" smoking among f~mzl~ sub~cc~ who never smoked. H~ refrained z~cmp~ing to do so for ~h~ following reasons: Sinc~ his prospcc{ive study was not designed for tha~ pu:- pose, no special information on ~h~ subject was ob- tained. Informa{ion was available on tb~ habits of the husbands of many of {he married w~men in lh~ study; bu~ not on lhc smoking hzbhs of former husbands of wom~n who w~rs widowed, divorced, separated or married for ~h~ second ~im~. More imporlant, in America ~ ~hat lime, women no{ generally barred from public and social ~ngs where men were smoking; and working husband~ who smoked generally did much if nor mos~ of smoking away from home". Dr. Hammond and Sslikoff then go on to poin~ ou~ why sludi~s in and l~pan ar~ mor~ likely to yield a m~aningful b~caus~ of ~he ~radilional segregation of marrie~ women from m~n o~hcr than ~h~r husbands in ~hos~ societies. Aher {he lung cancer work was publish~ (Oa~finkel 19gi) i~ soon bccam~ evid~nl, as mend had prcdiclcd, lha~ the s~udy had serious weak. n~sses. The American Cancer Society th~n probably d~cidcd lhal publishing ~h~ hcarl dala would be even less meaningful since ~he passive-smoking hear{ live risks are lower than the lung cancer risks, Th~ dif~cul~ics o~ carrying out m~an[ngful siv~ smoking s~sdies in ~hs Uni{ed Sa~cs, or ~urop~, ~ cmph~izcd by ~hc dala of Cummings (1990}. Hs found, in Buffalo, New York, ~ha~ nonsmokin~ women whose husbands smoked had urinary l~vcls only ~ higher (I0.~4 ng/mL vs. 6.7g ng/mL) ~han wom~n whose husbands did no~ smoke. indicated h~gh b~ckground l~vel of ETS exposure m~ans that i~ is very difficul~ to go{ a significant cffccl of passive smoking on lung cancer in ~he U.S. unless ~his background effec~ is mksn account. My current cstlmalc of rcla~iv~ risk for cancer from passive smoking based on ~he sludics on females is 1.17 afzsr corrcczion smoker misclassifica{ion. Co~ec~ing for background. using C~mmlngs' dala, increases ~his ri~k ~o 1.7. I~ E]ATCo document for Mayo Clinic 27 March 02

Letters to th~ ndhor interesting that, even ~in the West, studies can be carried out with tow bias corrections if the smoking prevalence, and hence the background, is low. For example, Trichopuulos etaI. (1983) and Kalaadidi etal. (1990) in Greece, where the respective female overstocking prevaleuces were only 10 and 17% and where married women don't mix with men other than their husbands, both found a passive lung cancer risk of :2.1o and Butler (1990), among female Seventh Day Adventists in California where the eversmoker prevalence was 14%o found a passive lung cancer risk of 9_.0. For all three of these studies the calculated bias due to smoker miselassifieation is very low, 0.2% or less. In summary, there has been no good reason so far advanced that warrants discarding the epidemiologi- eal results on passive smoking. Particularly, there is no good reason to discard them in favor of simplistic dose approaches such as those advanced by Mr. Lee. The best available risk assessments for passive smok- ing deaths in the U.$. continue to be those such as Wells (1988) that are in the 50~O/y range. A. Judson Wells 41 Windermere Way Kennett Square, PA 19348 REFERENCES Bond, LA.; Dutch¢r, I.S.; Mcdinsky, M.A.; Henderson, Birnbaum, L.S. Disposition of 1412 methyl bromide in rats after inhalation. Toxicol. Appl. 9harmacol. 78:259-267; 1985. Bond. LA.; Sun, I.D.; Mcdinsk~. M.A.; I~nes, R.K.: Y~h, Deposition, mctsbolism, and e~fion o~ 1-t4¢ nitropyren¢ and 1-14c n[tropyren~ aoated on dles~l exhan~t ptniele~ ~nfl~ene~d by ~xpo~urn eene~ntr~tion. To~i. Appl. Phar- macoL 85:102-117; 1986. Brown, D.C.= Pc:clan, ~; Garner, I.~. Can==t of ~= ¢~i~ sad =moking ha=band. Can. ~=m. Physician 2~:499-502; Buckicy, LD.; H=rri=. R.W.C.; Doll. ~.; V===fy, M.P.: Wil- llam=, P.T. C~==-control =lady of the husbands of women with dysplasia or carninom• of the ce~iz ut¢~. Lancet fi: lOIO- Budcr. T.L ~c rcl=fionship of p=ssivc smoking m various hcshh outcomes among Seventh Day Adventists in ~liromit. presented st ~ Seven& World Confcmnon on Tobacco and Health, P¢~, Australia, Apdl 1-~. 1990. In: Abstracts and pa~icipants. Pc~: Hcslt~ Dcps~cut of West¢m Austrtlis; 1990:316. ~mming,, K.M. Statement before Scientific Adviso~ Board, Indoor Air Qu shay and Total Human Exposure Committcc. ~ronmtntsl Prot==~ea A~ency. Washington. D.C.. bcr 4, 1990. Gs~mkcl, ~ ~mc as=ads ~ lung cancer mortality among non- smokers and = not~ on passive smoking. L Nat. Cancer Inst. 66:1061.1066; 19~1. Olan~, S.A.; P~lcy. ~ Passive smoking and hcs~ d~sesffie: cpidcmiology, physiology, tad biochcmht~. Circulation 83; 1- 12; 1991. Grimmer. G; Bruno. H.; Dettbara, G.; Naujack, K.-W.; Mobs. U.; Wcnzcl-l[artung, R. Contribution of polycycl~o aromatic com- pounds to the carcinogenicity of sidestrenm smokc of cigarettes evaluated by implantation into she lungs of rats. Cancer Loll. 43:173-177; Hsddow, I.E.; Palomakl, G.E.'- Knight. G.L Use of scram cctinine to assess the accuracy of self-reported non-smoking. Br. Mcd. L 293:1306; 1986. H•ddow, I.E.; Knight, G.J.; Palcmaki, G.F..; Kings, E.bf.l Wald, N.J. Cigarette consumption and serum cofinine in relation to blnhweight. Br. J. Obrtet. Gynaccol. 94:678.-681; 1987. Hsddcw, I.E.; Knight, G.I.; Palom•kl, G.E.: McCarthy, Second-trimester scr~m cotinlne levels in nonsmokers in rela- tion to bir~h weight. Am. J. Obstnt. GynccoL 159:481-484; 1988, Hammond, E.C4 Sclikoff, I.J. Passive smoking and lung cancer with ¢~mmcnts on two new papers. Environ. Rex. 24:444-452; 1981, Hcllbcrg, D.; Valantin, 5.1 Nilssen, S. Smoking as • risk factor in ¢¢rvlcal neoplaeia. Lencet ih1497: 1983. Henderson. R.F.; Bechro!d, W.E,; Medlnsky, M.A.; P.-Fischer, L; Lee, T.T. The effect of molecular weight/Iipophilichy on clearance of organic compounds from lungs. Toxlcol. Appl. PhnrmncoL 95:515-521; 1988. Kslandldi, A.; Katsouyanni. K.; Vovopoulnu, ~.; Bastas, G. Sarac- ci, R,; Tdohopoulos, D. Passive smoking end diet in the etiol- ogy of lung cancer among non-smokers. Cancer clts$¢l and Control i: 15-21; 1990. Khout'7, M.L An estimate of adult mortality in the. United States from passive smoking; • response to criticism. Environ. Int. 17:3St: 1991. Koo. L.C.; Ho, J.H.-C. Worldwide epidcmiologicaI patterns of lung cancer in. nonsmokers. Int. L EpidemioL 19:S 14-231 1990. Lee, P..N'. Passive smoking and lung cancer association--a result of bias'/Hum. Texieol. 6:517-524"- 1997. Lee, P.N..Mi$classlficatinn of smoking hahhs and passive smoking. A tcviow of the evldenc¢. Heidelberg; Springer-Vedag; 1985. Lot P.N. Passive smoking and lung cancer:, fact or fiction. Bicva. C.J.; Counois, Y.; Govserts. M, cds. Present and future of indoor sir. Amsterdam: Excerpts Medic•: 1989:119-129. Lee, P.N. An estimate of adult mortality in the United States from passive smoking (letter). Environ. Int. 17:379-381: 1991. Prhchard, LN.; Black. A.; McAughcy. I.I. The physical behavior of sldestream tobacco smoke under ambient conditions. viron. Tcch. Left. 9:545.552; 1988. Sandier, D.P.'. Everson. R.B.: Wilcox. A.J. Pasdvc smoking in adulthood and cancer risk. Am. J. Epldcmiol. 121:37-48: 1985. Slattcry~ ~.L. ctal. CigarcU¢ smoking and exposure to passive smoke arc risk factors for cervical cancer. L Am. Mcd. Assoc. 261:1593-1633; 1989. Trlchopoulos. I3.: Ksl sndidi. A.: Sparros. L. Lung cancer and passive smoking. Conclusion of the Greek study. Lancet, i~:677-678; 1983. USEPA (U.$. Environmental PrOtection Agency). Health effects of passive smoking: Assessment of lent cancer in adults and respiratory disorders in children. Review draft. Washington: U.S. Environmental Pmtectlon Agency: 1990. Wells, A.J. An cstlmate of adult mortnlhy in the Unltcd States from passlvc smoking. Environ. Int. I4:249-26S: 1988. Wells. A.L An estimate of sduh monalhy in the United States from passive smoking; a response to criticism. Environ. Int. 16:187-193: 1990a. Wells. AJ. Smoker misctassificatlon does not account for ob- served passive smoking ~sk for lung cancer. Submission to ScientiSt Advimry Board, Indoor Air Qu ality and Total Human Exposure Committee. U.S. Environmental Protcctlon Agency. Washington. D.C.. December 4. I~90~. Welts, AJ. Breast cancer, cigarette smoking, and passtve smok- ing. Am. L BpidemioL tlg:20S-2t0: 1991. BATCo document for Mayo Clinic 27 March 02

396 OBSERVATIONAL VS EXTRAFOLATIVE MODELS IN ESTIMATING MORTALITY FROM PASSIVE SMOKING Dear Editor: Lee (199I), in his reply to our criticism (Rcpace and Lowrey 1990) of his negative views on passive smoking and lung cancer (Lee 1990), now concedes that som~ effect of environme, ntal tobacco smoke (ETS) on lung cancer among nonsmokers is plausible. However, Lee disputes our contention that there is =remarkable agreement" ~ong risk ass¢ssor$ using di£f¢rent approaches to cstimatt the Hs~ of passive smoking (R,pac¢ and Lowrey 1990a. 1990b). To th, con~y, Lcc asscrts that ~¢r= is a ~massiw dis- crcpancy~ bctwccn risk ratios for lung canc,r d¢rivcd from passive smoking cpid¢miotogy and risk ratios "dosimc~ically" cstimatcd by ~xtrapolation from of- fccts in smokers to those in nonsmokers. A discrcpancy cxis~ only if one's prc-conc¢ivcd notion is that the dose-response relationship tobacco-smoke dosc and lung cancer must b¢ linear over ~r¢¢ ord,rs of magnitude. Lo~ implicitly as- sumes such linearity, but offers no ¢vidonc¢. Th¢ro is no in~insic rcason for biological phenomena such as c~cinog¢ncsis to bc linear ov¢r sev¢ral ordcrs of magnitude of cxposurc. In fact, nonlinearities twccn cigar¢ttcs smoked p~r day and lung cancer rat¢s ar¢ evid¢nt in studies by both Hirayama (1974) and Doll and P¢to (1981) cv¢n over the single order of magnitude dose range between light and heavy active smoking (Rcpac¢ and Lowrcy 1985). Let us compare ~0 predictions of hypothetical linear and nonlinear dosim¢~ic models for the risk of lung cancer from passive smoking. This will test Let's h~othesis that dosimc~ic models ~, =massivc- ly discrepant" with ¢pidcmiology. Equation 1 repre- sents a lin¢ar form of dosim¢tric mod,l for risk R(d) as a function of dos¢ d, and Equation 2 represents an equally arbitrarily chosen non-linear mod¢t: R(d)'= R== d/D== (I) R(d) = R,,{I - C~v='~ (2) where R(d) is the excess risk (above background) from exposure to tobacco smoke (in units of lung cancer deaths (LCDs) per tO0 000 person-years) as a function of inhaled dose. For a background rate of B = 9.R== = 316 - 9 = 307 LCDs per 100 000 person- years, is the average risk of active smoking above background; Dr= is the dose from average active smok- ing: the average active smoker smokes 32 cigaret- tes per day and is estimated to inhale a dose of D,= = 544 mg of tar per day (at 1980 average mainstream tar levels); k is a constant arbitrarily chosen to be 10. (Other constants from Repace and Lowrey 1985.) Repace and Lowrey (1985) have timated that the most-exposed and average- passive smokers inhale doses d of 14 mg and 1.4 mg tar respectively from ETS per day. The ratio d/D== = :544/544 = I, for average active smoking; 14/544 = 0.026. for heavy passive smoking; and 1.4/544 = 0.0026, for average passive smoking. Let us now compare the predictions of these two models based upon the parameter values for d/D== given above. Equation l, the Iin¢ar extrapolation. predicts an excess risk of 8 LCDs per I00 000 person-years for heavy passive smoking, and 0.8 LCDs per I00 000 person-years for average passiw= smoking. By contrast, Equation 2, the non-linear extrapolation, predicts an excess risk of 72 LCDs per 100 000 person-years for heavy passive smoking, and 8 LCDs per 100 000 person-years for average passive smoking, which are nearly identical to the risk values calculated by Repute and Lowrey (1985). We now use the values derived from each model to compute the expected risk ratios for passive smoking. The risk ratio for passive smoking, p, may be calculated from Equation 3: p -- [(R(d) + B)/B] (3) For the non-linear extrapolation model for passive smoking, p = [(8 + 9)/9] = 1.9 for the average case, (similar to the value of 1.8 found by Hirayama (198 l) in his cohort study of 91 540 female Japanese passive smokers). By contrast, the linear model predicts a risk ratio of p = [(0.8 + 9)/9] = 1.09 for the average case (about what Le¢ found in his own case-control study of 168 British passive smokers). Thus, it is clear that the predicted risk of passive smoking is a function of the choice of extrapolation model, and, contrary to Lee's assertion, dosimetric extrapolation models are not massively discrepant with observa- tional data from epidemiology. In the abs¢nce of experiments to determine the shape of the dose-response curve, the only way to validate any model is to compare it with observable human epidemiological data. We chose to reject a linear dosimetxic model beeaus¢ it was inconsistent with the bulkof human observational data. Lee chooses to accept a linear extrapolation model despite its inconsistency with observational data. Row else would a proposed risk assessment model be validated other than by comparison with human observational data? We do not favor extrapolation models in general because of the uncertainties introduced by extrapo!a- Co 3ATCo document for Mayo Clinic 27 March 02

Letters to the cdRor tion over several orders ot" magnitud~ from high ~x- posuros in smokers to low exposures in nonsmokers. Rather, we chose a phenomcnological approach r~.lat- ing cnv~onmcntal risks in nonsmokers r~]al~d ~nvironm~n~ ~xposurcs of nonsmokers. Wc validated our mod~t by predicting--to within 5%--the risk ratios and ~k rat~s obliged in U.S. cohort an~ 198~. By con~ast. L~ h~s unconvincingly to oxplain away th~ o~semational ~ata by a mathemati- cally po~bl~ but physically improbabl~ explanation for lung can~ morality ratios from passivo ~mok- ing cpid~miology. Under L~o's hypothesis, th= rat~ o~ smokers mis~Ia~ffi~d as nonsmokers diffcm sig- nificantl~ ~rom Ash to North America, an~ ~ot (implausibly) of thc exact magnhu~ on bmh con- tinont~ aa~ in oath ~tu~y as to produco ~imil~r but "spurious" positive results (od~s ratios clustered tween 1 and ~) in more than ~wo dozen epidemiological studies around the globe. Similarly, Lec also unconvincingly attempts to plain away the virtual absence of negative studies passive smoking and hear¢ diseas~ moreality by turning without proof ~hc existence of publica~oa bias, ignoring the ove~h~lming biological plausibility for ~ ~f~t (~ non-lin¢~) ~ demonstrated by ~lan~ an~ Parmley (1991). Lee expects us to accept that ~horo has b~cn publication bias in ehe Amebean Can- c~r Society (ACS) study, on the basis of his un- found¢~ belief that th~ £CS looked into heart disease morality, found no association, and therefore did not publish. Since there was no data-base from which to draw conclusions in this area, we created one. Ac- cording to Lawrence Garfinkel, recently retired President for ~pid~miology of ~e ACS, th~ ACS had in hot n~ver ~xamin~ its data for heart dis~as~ mortality from passiv~ smoking. Finally, we would lik~ to observe ~hat our own calculations of lung cancer Dora passiv~ smoking, publishe~ six years ago in this journal, ~sdmat~d a mortalhy ratio of 1.% adjusted for workplac~ passive smoking, for the ACS cohort studi¢~ by ~¢pace and Lowrey 1~85). This is in agreement with ~ recent meSa-analytic valu~ of 1.~ calculated by W~lls (1991), adjust~ for non.domestic ETS posures, based on new studies of cotinine levels in U.S. nonsmokers. James L. Repace U.S. Environmental Protection Agency Washington, DC 20~60 and Alfred H. Lowrey Naval Research Laboratory Washington, DC 20375 Dolt. g.; Pete, R. The causes of cancer, O:cford University Press. New York, NY; 19gt. Giants, S.A.; Psrmley, W.W. Passlvc smoking and hcsn disease: cpidemioloSy, physiology, and blochcmist~. ~ircolation 12; 1991. Hirsyama, T. Prospccfiv* studlcs ca can~cr cpidcmiology has~d on con,us population ~ lapan, h: Prec. XI ~tcm~donal Can- ccr ~ongr~ss, Florence; ~0-26 O~obcr, 1~74. ~zcc~ta Mcdlcs, Amsterdam; 1975:26-35. ~tyama, T, Pa~ive smoker and lung c~cer. Br. Me~. ~. 282: ]93- 1~4. L~e, P.~. An e~dm~e of adult mo~ai[ty ~ ~e United State~ from passive smokeR; a tc*pousc Octtcr). ~nviron. ~t, 16:17~-1 1990. L::. P.~. An esHmate of adult mortality in the United S~tcs from passive smoking 0eater). Environ. Int. 17:379-381; 1991. geptce, LL.; Lowtey, A.H. Aqutnt{tative estimate of nonsmokers' lung cancer risk from passive smogs. En~ron. Ins. ! 1985. Repae¢. I.L; Lowrcy, A.H. A rebuttal to c~ti~sm of ~c phenome- nologlcal model of nonsmokers' lung cancer ~sk from passive smoking. ~nviron. Ctrcino. Revs, I, Environ. Scl. Heal~ pt Cf4):225-235; 1986, gtpace, J.L; ~wtey. A.H. Predic~ng ~e lung c~cer Hsk from domesti¢ passive smoking. American Rcv. Reap. Dis, 1~6:1308; 1987. Repute. I.~; Lowrey. ~.H. A Rebuttal to ~e~t~nstcin common. ta~ on passive ¢moklng ~k ~¢ttcr). Enrich. ~t. 16:183.184; 1990a. geptcc. LL.; Lowrey, A.H. Risk assessment methodologies in ptssi¢e smok~I. Risk ~al. 10:27-37; 19~b. Wells, AJ. An estimate of adult moatHty M the United from pttsiw smokh$: t response to criticism 0case0. ~nvlroa. Int. 17:382-385; 1991. The opinions expressed in this Ietter are those of the authors, and do not necessarily represent the officlal policies of their r::tpcotlve federal agencies. BATCo document for Mayo Clinic 27 March 02