Tobacco Documents Online

~rn•iaonmmr lnrernorronal, Voll I l,,pp, 3-22, 1985 Prtnted in the USA. Atl rights reserved. A QUANTITATIVE ESTIMATE OF NONISMOKER'S'' LUNG CANCER RISK FROM PASSIVE SMOKING J. L. Repace U S Envtronmentat Protecuon Agency. Wasnmpton, CJC 20a60. IDSA• A. HI Lowrey Naval Researcn Laboratory. Nrashrnpton, DC 20375. USA" (Reeeived, 17'.Wo) 1984; AAccepted 11 January 1983/ This work presents a quantiutivrassessment of nonsmokers' risk of'lung,cancer from passive smoking. Thr,esnrnates given should be viewed as prrlimihary and isubjectto change as improved research becomes available. l.t is estimated'that t),S~ nonsmokers are exposed to from 0 to 14 mg of, tobaceo tar per day,;, and that the typical nonsmoker is exposed to 1_4 mg per day,. A,phenomenologicai exposure.response, relationship is deri.ed, yielding 3 lung cancer deaths per year per 100,(100 ipenons exposed, permg daily tar exposure. This relationship yields lung cancer mortality rates and mortality ratios for a U:S! cohort ..hich are,consistent to wvhin 5~ro with the results of both of the large prospective epidemiological studies of pusisr smoking, and lung caneerin the UAited States and Japan: Aggregate exposure to ambient tobacco smoke is estimated to produce about SID00, lung cancer deaths per year in U.S, nonsmokers aged t 35 yr, with ian avenage loss of life expectancy of 17 a 9 yrper, fatality. The estimated risk to the mosttxposed passist smokers appears to be comparablrto that from pipe and cigar smoking. Mortality from passive smoking is estimated!to be about,t+vo orders of magnitude higher than that estimated for carcinogens currently regulated as hazardous air pollutants under the federallClean Air An. Iinrtroductiion Exposure of nonsmokers to indoor air pollution from tobacco smoke (alsolknown as involuntary or passive smoking) has recentl~- become a public health concern (LSSG, 1982) for several reasons: such exposure is widespread (Repace andlLowrey, 1980; Friedmani et al:, 19$3);,studies of the effects of tobact:osmokeon smokers worldwide have implicated it as the most important cause of lung cancer, (USSG. 1982: Doll and Pete; I'98'1); existcnce of a threshold for carcinogenesis is doubtful (USSG 1982; IRLG„ 1979; U.S. EPA, 1979a; IAR'C; 19?9; Pitor, 1981); and'.there is suggestive new evidence of lung cancer (and other serious health effects) in nonsmokers exposed to ambient con cent rations of tobacco smoke (Trichopoulos, 1981, 1!983;, Hirayama, 1981a, 119891b; 1983a, 1983b: Garfinkel, 1981; Correa et aJ:, 1983; Knoth et al:, 1'98'3„ Gillis et, al.,, 1983i Koo et al:, 1983: Kabat and' Wynder, 1!981;, luliiller, 1'984; Sandler er al.,, in press a„b). There are three important fractions of'tobacco 'The views preaenied in th s article ate those of the authors, and do not netiessarily reflea the poltaes of their respe.rtvc agen4ies. o160-a1C0"85ia.00 - .00 Copyright r 1985 Pergamon Press Ltd. smoke:, mainstream smoke, which the smoker inhales directly into the lung; exhaled mainstream smoke, that fraction of the mainstream smoke which is not retained'in the lungs of the smoker, and sidbstream smoke, thart fraction of tobacco smoke emanating directly from the burning end of the cigarette into the air; Nonsmokers are commonly exposed to tobacco combustion productss in diluted' sidestream and, exhaled' mainstream tobacco smoke from cigarettes, cigars, andi pipes (Repace and. Lowrey, 1980): Tobacco smoke contains 60 knoscn or suspect carcinogens, including 5'l' in the phase contain- ing particulate matter; the carcinogenic actir,its• of to- bacco smoke appears to require this phase (USSG. 1982),. Animal bioassays indicate that siidestream to- bacco tar is more carcinogenic per unit w,eight than main- stream: tar ('H'.ynder and Hoffman, 1'967). For public health purposes, it will be assumed that mainstream and sidestream smoke have similar human carcinogenic potency. In his 1982 report on cancer and smoking (l,'SSG„ 1'982)„ the U.S. Svrgeon General asserted that: despite the incompleteness of the evidence, nonsrnokers should 3 AA-1

i' avoid exposure to~second-hand smoke to the extent possible, a rusk-management judgement supportedl by the World Health Organization andl the National Academy of Sciences ('WHO„ 1979; NRC, 1981). This raises the question of whether the' quantity of tobacco tar to which the'average nonsmoker is exposed creates a significant risk of lung cancer. In order to answer this ~ question, a quantitative risk assessment is first justified and then performed. R'i'sk assessment is the use of science to define the health effects of expo- sure of individuals or populations to hazardous materials or situations (NRC, 1983): Risk assessments contain some or all of the following four steps: (1')! Hazard identification-the determination of whether a particular chemical is or is not causally linked to certain health effects. (2) Dose-response assessment-the determination of therelationi between the magnitu'de of exposure and the probability of' occurrence of the~ health effects in ques• tiom (3) Exposure assessment -the ' determination of the extent of human'exposure before or after application of regulatory controls.. (4) Risk characterization- the ' description of the nature and' often the magnitude of'the human risk, inL cluding attendant uncertainty,. In other words, quantitative risk assessment deals with the question of how much morbidity and mortality an agent is likely to produce gii.•en specified levels of ex- posure. Typically utilized in the regulation of carcino- gens, it is important because control efforts cannot pro• ceed without assurance'that the health gains are worth thecost(Lave: 1983; Albert, 1983). Onthe'basis of such assessments, informed risk management, judgements can, be made: This work draws upon the epidemiology of lung cancer (USSG. 1982; Pitot, 1981; USSG, 1979; Ives, 1983) and on indoor air pollution physics (Rrpace and Lowrey, 1980, 1982; NRC, 1981) to produce a risk anal, ysis (IRLG, 1979; U:S: EPA, 1979a; Lave, 1!983; COST, 1981; Fischoff et a1- 19'81; NRC, 1983) in'which non- smokers' lifestyles are correlated to exposure to airborne tobacco tar, and incidence of lung cancer. This analysiss first reviews estimates of the average exposure of'the' general population of ambient, tobacco smok~e: Second, it, reviews studies linking,tobacco-rel'ated'disease in non- smokers to exposure-related variations in lifestyle. Third, it couples these two factors to, develop al phe- nomenological! estimate fbr the' aggregate lun'g, cancer risk to the lJ.S. nonsmoking population, and develops an exposure-response relationship for the estimation of the risk to the most-exposed. Fourth, it compares the. estimated level l of lung cancer mortality and resultant loss of life expectancy. from passive srnoking to thosee from cigarette, pipe, and cigar smoking. Fifth, it com- 3. L. Repaceand.C. H_ L.)..nea, pares the predictions of alternate exposure-response' relationships with the results of two large prospective'. epidemiologic studies of passive smoking and lung', cancer, and performs a sensitivityy analysis. Finally this: work compares the estimated risk from ambient tobaccoo smoke to th2t from various airborne carcinogens cur- rentJy being rtigulated in the'United States as hazardouss air pollutants, to place the significance of the estimated' risk, in perspective. Variation of Exposure with Lifestyle In earlier work (Repace and Lowrey, 1980, 1198P., 1'983, 1!98!4;, Repace, 1981, 1!982, 1983, 1984, in press; Repace er al., 1980; 1984; Bock er aL, 1982)factors af- fecting nonsmokers' exposures to tobacco smoke were studied, and field surveys of the levels off respirable par- ticles were conducted' indoors and out, in both smoke- free and smoky environments. This work established that ambient tobacco smoke imposed significant air pollution burdens on nonsmokers, and, using ct)ntrolled experiments (Repace and Lowrey, 1980, 1982, 1983'), a model was developed to estimate those exposures. This model predicts that the exposure of U.S'. nonsmokers ranges from 0 to 14 mg of cigarette tar, per day (mg/d), depending upon the nonsmokers's lifestyle: As derived in Appendix A and showninTable L, the average popu+ lation exposure for adults of working age, averaging over the work and home microenvironments, is about 1.43 mg/day (Repace and Lowrey, 1983) with an 861r0 exposure probability. Table 1, derived from the model' in Appendix A, estimates probability-weighted exposure toithe paraicu- late phase of ambient tobacco smoke for a typical U'.S. adtdlt nonsmoker. Exposures received in other (Repace er al:. 1980) indoor microenh•ironments, outdoors, andd in transit, which account for the remaining 12ro of peo- plt:'s time, were omitted. Table 1 is derived from con- sidtrations that ambient concentrations of~ tobacco tar have been found to be directly proportional to the smoker density and inversely proportional to the ven- tilation rate (Repace an'di Lowrey, 1980): Ventilation rate tables given by ASHRAE (1!98'1'), can be used to estimate both the range in, ventilation rate, (from thee design mechanical rates) and smoker density (from thee design occupancies), and thus upper and lower bounds an'dI average concentrations for' model workplace and, home microenvironments can be es'timated.. Table I suggests that individuals receiving exposure both at home' and at work constitute a high exposure group,, with the work'place appearing, four times as strong alsource of'exposure as the home; the reason for this di'fterentiall is the generally higher occupancy (i.e.,, smoker density), encountered in the workplace (Repace and Lowrey, 1982, AS!HRAE; 19811). This estimate of' exposures represents a modeled weighted average taken over the entire population; including thosrwfio!are not exposedl /lA'-2'

:7an:er rtsk from,passi~e smuktng, Tlabie,l. Estimated probabilities of nonsmokers' exposure to tobaeco smoke at home and,at work (aiten Repare and!Lnw•rey, 19831 Appendta: A):+ Monexclusi%e probabAitp of being,exposed at work: 6?6'e; probability of'not being exposed at work: 37%. Nonexclusive probability of being exposed, at home: 620'ro; probability of not being exposed at,home: 380".,. L ifestyle: Daily Average Probability of Heing Exposed (Rounded Valites) Modeled Daily Average Exposure (rng) Daily. ProbabifityWetghted At work and at home: r. 63 x 62 w 39 2.27 0:89 Neither at work nor at home: ra 37 x 38 - 14 0.00 0100 At home but not at work: 40 62 x 37 : 23 0.45 0:10 At work but not at home: % Touli 2% 63 x 38 = 24 tIOiD t.82 0144 t.a3 +The estimated exposure to the paniculate phase ofiambienr tobacco smoke,for 1U.S' adults of working age, at work and at home (these two:micrioenvironments aecount:for an estimated 880,16 of theatierage person's-both smokers and nortsmokers:-time); determined fzom:average cotueentrations of'tobaeco smoke caltulated for model workplace and hotne taicroen.ironmenas, weighted for average occupancy, as derived in Appendix A. Jarvis and Russell (in press), in a study of urinary cotinine (a nicotine metabolite) inia sample of,1211 self- reported nonsmokers, state that only 12e%o of'subjects had undetectable cotinine levels, despite nearly 50t^o reporting no passive smoke exposure. In a study of 472 nonsmokers, ]Wfatsukura (]984) examined the relation. ship of'~ urinary cotinine to the smokincss of their en, vironment4 and found that nonsmokers who liivedl or worked with smokers had higher cotinine levels than those who did not. Matsukura et at:, (1984) also found that cotinine levels increased with the numberof'smokers present in the home and the workplace; however, none of the differences were statistically significant, except for the lowest urinary cotinine level of'the nonsmokers who were not exposed to tobacco smoke in the home or the workplace. These studies respectively illustrate the widespread exposure of nonsmokers to ambient tobacco smoke, andltherelative importance of'the domestic and workplace microenvironments in such exposures. Epidemiolbgical Evidence forthe Variartion of Risk with LiKestyle: Pultmonary Effects Vk'hite and Froeb (1980) evaluated, the effect of various degrees of long,term (>20 N•r)~workplace expo- sure to tobacco smoke on, 21,00 healthy middle-aged workers, Of these workers,, 89'O•'o1 held professiional; managerial; or technical positions, while the remaining 170'o were blue-collar workers. Relati% e to those npnex+ posed anhome or at work„passive smokers of both sexes suffered statistically significant declines in mid', and end-expiratiory flow rates which averaged about 1315Pio and .207o respectively, andidid not diflfer significantly from,the values measured in noninhaling or lighrsmokers of cigarettes, pipes, and cigars: Tlhe> concluded that chronic expcnsure to tobacco smoke in the work environ= ment, reduces smalll airwa5ss function t~o the same extent as smoking ll to 10 cigarettes per da%: S Kaufiftnartn et cl: (I1983) compared! puhnonary func- tion in about 3800 people in France: 849 male "true" nonsmokers (defined as those not exposed at home), 165 male passive smokers (def;ned' as those exposed' at home)„ 826 female "true" nonsmokers, and 1941 female passive smokers. The authors restricted the analysis to subjects aged'40 yr orolder (i.e.. to those,who had been exposed for 15 or more yearsto smoking by their spouses) and who were living in households with no persons over the age of, 18 yr except their spouses. They found thatt nonsmoking subjects of'either sex whose spouses~ were currenn smokers of at least 10 g (about 10 cigarettes) of tobacco a day had mid-expiratory flow rates averaging 11.5°'0, lower than, those married to nonsmokers. For women in social classes with the highest percentage of paid work, the effect of workplace smoking appeared to confound the effecrof passivesmoking,at home: How- ever, in the large subgroup of women without paid work (iie., not exposed to workplace smoking), a!clear dose- response relationship tb amount of husbands' smoking was observed: They concluded that women, living with hean•y smokers appeared to have the same reductions in mid'•expiratory, flow rates as light smokers, an& that after 15 yr exposure in the home ervironment, passive smoking reduces pulmonary function. A third'.study by. Kasvga (1983) of' urinary hydroKy- profine-tio-creatinine (HOP-r) ratios as a function off passive smoking stiatus showed that HOP-r levels in nonsmoking wives and childremvaried in aidose-tesponse relatiionship, %vith husbands' and parental smoking hab- its, when adjusted for pre-existing respiratory disease. }:asugal(1983) asserts tharHOP-r serves as a marker to derect, deleterious active and passive smoking effects on the lung, before and after the manifestation of clinical symptoms, and' that urinary HOP-r in light-smoking womeni is almost equivalent to HOP'-r in nonsmoking wives with heavy-smoking husbands.. These three epidemiologic studies provide evidence that variations in the exposure of adult: nonsmokers to ambient tobacco smoke at home and, particularly, at 0

6 work, can produce observable pultnonary effects. Like effects have been observed in children exposed at: home (Tager et at.,, 1983). Cancer Thirteen epidemiolbgic studies have explicitly ex- amined the lung,cancer risk incurred by the nonsmoking spouses of cigarette smokers. In all but one study„ thee only exposure variable was the strength of the spouse'ss smoking habit~ The studies were conducted in Greece (Trichopoulos et al., 1981, 1983), Japan (Hirayama„ 198'1a, 198'1'b, 1'983a, 1983b), the United' States (Gar- finkel, 1981; Correa, et al., 198'3: Kabat, and Wynder, 1984; Miller, 1984; Sandler, et al., a and b, in press),,, Germany (Knoth et a1:, 1983), Scotland (Gillis et al:, 1'983), and Hong Kong, (Chan and Fung, 1982; Koo et aL, 1983). In the Greek study, Trichopoulbs et al: (1981, 1!983) used the case-control technique: involuntary exposure to cigaretrte smoke as measured by the husbands" daily consumption was found to increase the average risk of lung cancer, by a factor of 2.4 (p < 0.011) when 77 lung cancer patients were compared to 225 controls, and a dose-response relationship~ was observed. Divorce, remarriage, husband's death, and change in smoking habits were considered. In the Japanese study (from 1966: to 19811) of lung cancer in 91,540 nonsmoking women, Hirayama (1981a; 1981b,, 1983a„ 1983b) used the prospective technique: relative to those women not exposed arhome (controls), involuntary exposure of wives of'smokers was found to increase the average risk of lung cancer by a factor, of' 1.78 (p < 0:001)i where the exposure was also estimated from husbands' daily consumption. The an- nual lung, cancer death (LCD) rate in the controls was 8.7 per, 100,000: Hirayama found that the exposed wives experienced an average annual increase in lung, cancer mortality rate of 6.8 per 1100,000, with a range of 5.31 to 9.4'.per 100,000, in aidose-response relationship depend- ing upon the degree oflthe husband's smoking. Hinab•amaa foundi furrther that the risk of lung cancer death in non- smoking women, increased both with the time of expo- sure and number of cigarettes smoked daily by the hus- band. Hirayama also reported a factor of 2.9 ( t 0.3', at the 9S'eo confidence level) for increased risk of lung, cancer in 1010 nonsmoking husbands with smokiiag, wIbY5,. More recently, Hirayama extended his earlier work tio; suggestl increased risk of nasal sinus:cancer, andlischernic heart disease in passive smokers, andi evidence of de- creased lung cancer risk in nonsmokiiog, wives of ex- smokers. With respect to cancer of the para-nasal si- nuses in nonsmoking wives (rt = 28). Hirayama foundl standardized mortality ratios of 11.00; 2..7, 2.56, andi 3.44 when husbands were nonsmokers, smokers of 1-l!4, 115-19; and >20,cigarettes per day, res~pectir;ely l. L. Repace and A. H. LowRey (p = 0.01). For ischemic heartl disease. risk elevations: for nonsmoking wives (n = 494) with the extent of' husbands' smoking were reported, with, standardizedi mortality ratios of 1.00, 11.10, and' 1.31 when; husbands were nonsmokers,,srrtoken of 1-19; and s20 cigarettesl day, respectively (p c 0.02). For lung,cancer, the stan- dard#zedlmortality ratio1 of lung cancer in.nonsmoking women (n. = 200)l was 1.00, 1.36, 1.42, 1.58, and 1.91 when husbands were: nonsmokers, exsmokers, daily, smokers of 1,14, 15-19, and >20 cigarettes/day,, respectisely: In the first U.S. study, Garfinkel O98I1) reported results from an analysis of' data collected' from the American Cancer Society's (ACS) prospective study of' lung cancer risk in 176,739 nonsmoking white women (1'960 1to 1972) as a function of involuntary exposure as indicated by their husbands' cigarette consumption. Of the total„72470 of'the nonsmoking women were married to smokers. Three smoking categories were identifi;edz none, less than, one pack (20 cigarettes) per day, or greater than one pack per day. Garfinkel reponed'statis- tically insignificant risk ratios of 1.00,, 1.27, and 1.10,, respectively, for the three categories (average 1.20'over the exposed categories). AlSo reported were age-stan- dtzrdized death rates, which were respectively 13.8, 12.9, and 13'.1 lung cancer deaths per 100,000 person-yr for this cohort in, 1960-1964, 1964-1968, and 1!968-1972 (averagc 13.3' per 100,000 person-yr for the period 11960-1972). The death rates were standardized', to the distribution of white men and women combined for the U! S. population in 1965, which decreased the rates for females "slightly." More recently, Correa et a1. (1983) studied 8'maleand 22 female nonsmoking lung cancer cases and 180 male and! 1'33' female controls as part, of a larger study includ- ing smokers, with.1'338 lung cancer cases and 1393 con- trols, in Louisiana; They reported that nonsmokers married to heavy smokers had an increasedirisk of lung cancer~, as did smokers whose mothers smoked. Men with smoking wives had a nonsignificant risk ratio of 2.0 compared to their counterparts with nonsmoking wives, and women with smoking husbands had an antrage risk ratio of 2.07 (p < 0:05) compared to women with nonsmoking, husbands. An exposure-re- sponse relationship was observedL with the peak risk reaching 3.5 (p < 0.05), The combined data for men and women, passii.•e smokers was significant (p < 0:05) for the heaviersmoking category ('a4111 pack-years), A third U.S'. case-control study, by Kabat and Wvnder (',198a), reported on passive smoking and lung cancer in nonsmokers for 25 malkcases and 25 controls, and 53 female cases and 53 controls, where the majority of the patients were from New Y'ork City. The controls consisted of patients hospitalized for non-smoking- related& diseases, roughly two-thirds being cancer pa- tients. No differences on exposure to passive smoking,at. home or at work were found in the women. H owe+•er:, AA-4

Cancerrisk from pauive',smokong the male passive smokers-di'splab•edla statistically. signifi- cant (p = 0.05) di'ffenence in lung cancer (odds ratio 1.6) relative to the non-exposed group. A fourth U.S. study by Miller (1984) of mortality from all forms of cancer in 123 nonsmoking women (only 5 lung cancer cases) as a function of' husband's smoking history reported a nonsigni'fiicant odds ratio of' 1.4 for all women (p = 0.15) for women whose hus- bands smoked relative to those who did not; when em- ployed' women were excluded, the odds ratio increased to 1.94 and was statistically significant (p < 0.02),. A fifthiU.S. studyof'Sandier et al: (in press a) aho ex- amined mortality from all forms of' cancer rclated' too passive smoking, in'both nonsmokers andismokers [231 cases and 235 controls (70% white and 67'eIo i female); only 2 cases of lung cancer ininonsmokers) as a function of spottses' smoking habits. Cartcer risk-adj,usted odds ratio-(lung, breast, cerviz, and endocrine) among in- dividuals ever married to smokers was 2.0 times that among those never married to smokers (p 1 < 0,01). This increased risk was not explainedl by confounding in dividual smoking habits, demographic characteristics, or social class. In a sixth US. study, Sandler et ol, (in press b) ex- amined' cancer risk iniadulthood in 197 cases and 223 controls, 66'% female, from earay life exposure to par- ents' smoking. They, found that mothers' and fathers' smoking were both, assoeiatedI with risk for hemato- poictic cancers (Hodgkin's disease, lymphomas,: and! leukemias), and! a dose-response relationship' was seen' for the Iatter two. The odds ratio for hematopoietic cancers increased from 1.7 when one'parent smoked, to 4.6 when both, smoked (p < 0.001).. lnthcfirst of two studies from Hong Kong, Chan and Fung (1982):found a lowerincidence of passive smoking among 34 female lung cancer cases (40:5°jo) than'among 66, female controls (47.5°l0): All patients and controls were interviewed concerning their smoking habits and those of, their spouses, their cooking habits, including types of cooking, fuel used. Histological diagnoses of tumors were obtained. Controls were t'aken from ortho- pedic patients. In the second H'ong Kong study; Koo eral. (1983) studied passive smoking in' 56 female lung cancer cases and' 85 female controls. Passive smoking cases had an excess of 3'.8 yr of, passive smoking (workplace plus domestic exposures) compared with controls, but tihee differences were nocstatisticallyt significant (p <0:069). However, among a su'bgroup~of'8 marine dwellers, cases had 1'1.8 years more exposure than controls (p = 0.0003): Knoth er al: (I'983) reported on a study of 39 non- smoking,German females with lung cancer,. 61.5'077o were found to have smoking spouses. The authors state that, this percentage was threefold!thatlexpected on the basis of smoking habits of Germanimales: Gillis er al: (1984) reported preliminary results of a study of passive smoking and lung cancer ini 91 male 7' controls (n = 2) [the numbers in parentheses give the' numbers rnfl cases] without domestic passive smok'ing and in 90 subjects exposed l at home' (n = 4); and in,40. femalt controls (n = 2) and 5,8 subjects (n = 6). No ef- fects of lung,cancer were noted in the femalt•s, but ele- vated rates of myocardial' infarction were reported (risk ratio 3''.0). In the males, elevated rates of both lung cancer (risk ratio 3,25) and myocardial infarction (riskk ratio 1.45) were reported: Gillis et aL state that since'inT sufficient time has elapsed since the' beginning of this study, no firm conclusions can be drawnrelating to the incidence of cancer, or other, diseases. Thus there are now a large number of studies provid- ing evidence for increased risk of'lung cancer from, in- creased exposure to passive smoking. It might be ex- pected that subgroups of the population which proscribe smoking among their membership would have a lower probability of passive smoking, and therefore should also have a lower incidence of smoking-relatedidisease than the general i nonsmoking, population. One such subgroup is the Church of Jesus Christ of' the Latter Day Saints, popularly known as the Mormon Church, which advises against the use of' tobacco. Ensuotn (1'978) found that active Mormons whoi were nonsmokers had standardized mortality rates for lung, cancer which were 21010 of those in the generali popula- tion which includes smokers. This rate was found com- parable to the rate of 19e/a for a sample of'the' U.S.. general population "who had never, smoked cigarettes." Interestingly, however, this result occurred despite the fact that 3l e/o of theactive Mormon cohort were former smokers. This confounding factqr, was not: present for certain subgroups in the following study: Phillips et al. (1980a, 1980b) have studied mortality. (from 1960 to 11976) , in Seventh Day Adventists (SDAs), a religious group that also follows rigorous proscrip- tions against the use of tobacco. As with the Mormons,, SDAs have rates of mortality fromi lung cancer and' other smoking related cancers that are fractions, 2'PaW and' 6fi"'o, respectively, of the rates for a demograph- icadlj• comparable group inithe general U.S. population (including smokers) (1'980a). M'any SDAs work for church-run bus'inesses; Thus, SDAs appear to, be less likely than the general population to be involuntarily ex- posed to tobacco smoke; as children or as adults, at home or in, the workplace, because neither SDA homes nor SDA businesses are likely to be places where smok- ing is perrniited„and because the great majority of SDA family and social contacts are among other SDAs whol do not smoke (See Appendix C): ared' mortality in l: Philli 11980a 1980b ( ps a ) ciomp el, N, two demographically similar groups of Southern Cali- CAI fornians: SDAs(from 1960 to 1976);and non-SDAs (from ~ 19ti0 to 1'97'11); A sizable.subgroup (35°ro) of SDAs re- ~ poot prior, cigarette use, especially among men ( I'980b'). ~ However, for two select subgroups of each group, 25,264 ~' SDAs and 50,216 non-SDAs who were self-reported AA-5

8' Table _. Age-adjustedlSDA-to-non4DA ratio oflun8 cancer mortality (after Phillips cr a!.' (1980b):, By Health Habit Index Average Best Third Awen`e, Third ~ Worst Third 1. All SDAs 0.54 OjSd 0:40' 0.9ti. 11. SDAs who never smoked 0.41 0:41 0.32 0.78'. Values shown iare adjusted'by'NtantelkHaenzel procedure (;p s' 0.0:1). •LunB cancer, mortality ratios taken from a,prospective study of'two demographically similar; cohorts. The non-SDA, come from the general south California population, and were: self-reported non- smo6ers who never smoked. The,SDA come from a southern Califor- nia subgroup less, likely to engage inn passive smoking by virtue of' Gfestp•le differences: The health lubit index is' a measure of' how faithfully individuals adhered to the Church's teachines; the worst third wwere alto more'likely to have rnon-SDA spouse: (Values quoted in text are the reciprocals of' numbers pven here:) Phillips er of.' (1!9b0a, 1981Db) reponed results for all SDA, and reported replicating these data for SDA who rKwer smoked. as shown. The SDr11 subjects and non-SDA subjects for, this study, consisted of white Californiaa respondents to the same four-page self-adtninistered questionnairee colleeted'by the American Cancer,Soeiety study of'I million subjects throughout ~ the United States (NCI, 1%6; Garfinkel. 1981;,Phi)lips er a1:,,1980a„ 1980b). nonsmokers who had never smoked, age-adjusted mor- tality rates were compared for smoking-related and nonsmoking-related diseases. Table 2' compares age- adjusted lung cancer' mortality ratios for two SDA co- horts relative to nonsmokers in the general! population who never smoked. The first cohort consists of'a)1 SDA, and includes those who neversmoked~ exsmokers, and smokers. The first row of Table 2 gives the mortality ratios relative to the never-smoked non-SDAs in the general' populatiom The second row compares the sec+ ond SDA cohort (those who never, smoked):to the non. SDA who never smoked. The values given are averaged over both sexes. F'rom,Table'2 the results show that the non-SDA group of nonsmokers who had never smoked (but who were more likely to suffer involunt'ary expo- sure to tobacco smoke) had an averageJung cancer'mor- tality rate of 2.4 times that of'the never-smoked SDAs (the group, less likely to have suffered such exposure by vii•tue of their lifestyle),. This concludes the review of evidence relating variations of lifestyle to variations in lung cancer risk in nonsmokers. Does Arrnbient Tobacco Smoke Pose a Carcinogenic Hazard? The International Agency For Research on Cancer ('IARC) criteria~ for causali'ty to be intlerred betnveen ex- posure and human cancer state that confidence in cau+ sality increases when (1) independent studies agree; (2) associations are strong;', (3) dosr-response relationships exist, and' (4) reduction in exposure' is followed by re- J. L. Repace and A. H. Lo~rev duction in cancer incidenea(IARC, 1979). These criteria are applied here as follows: 1'. There are now 14 studies; covering, 6 cultures, in- dicating a relationship between exposure to: ambient tobacco smoke and incidence of, lung canceri. If the studies are divided,into substudies of men and womt:n, this yields 20 substudies, all but 2'of which suggested an increased cancer mortality from' passive smoking; and 12 of which attained'statistical signi'ficance. Moreover, the mortality ratios ~based on spouses' smoking as an ex• posure variable, cluster around the': value' 2.0. Thus, many independent studies agree. 2. Mainstream tobacco smoke is strongly associated with lung cancer. The U.S. Surgeon General (USSG, 1982) asserts that mainstream cigarette smoke is a major cause of cancers of'the lung, larynx, oral cavity, and esophagus, and is acontributory'factor for the develop- ment of cancers ofi the bladder, pancreas, and kidney,, where the tetart contributbry factor does not excludethe possibility of' causality. Both smokers and nonsmokers are exposed to exhaled mainstream and sidestream tobacco ~ smoke. Sidestrearn smoke by animal bioassay has been found to be of greater, potiency, than main- stream smoke. 3'. Five of'the 14 studies reported dose-reponse rela- tionships between passive smoking and lung cancer. Dose-response relationships between lung cancer and active cigarette smoking show increasing mortality with increasing dosage of smoke exposure, and an inverse relationship to age of initiation (USSG. 1982). Dose- response relationships are also shown for smokers whose' smoking habits are like heavy passive smoking (Wynder and Goodman, 1983; Jarvis and Russell, in press), i.e:, in1 cigaretrte smokers who~do norinhale, and in pipe and cigar smokers, who also, are unlikely to ; in haVe'(USSG, 1982; USSG, 1979): 4. Reductions in:lung cancer incidence'for reduction in exposure have been.found in all majprstudies of'ac- tive' smoking (USSG,, 1982). The one study of passive smok'ing and lung cancer which examined this question also found a similarresulr(Hirayama, 1983b),. Further- more, the comparison of the SDAs who never smoked, and who should have reduced exposure relative to the non-SDAs who never smokedl alto appears to exhibit this effect. On the' basis of the IARC criteria„ the evidence ap- pears to be sufficient for reasonable anticipation of an increase in lung,caneer mortality#rom~passive smoking, justifying', a quantitative risk assessment. The signifi- cance of the public health risk willl now be estimated. Estimation ofiTo2al LCD Risk andi a Phenomenological Exposurtl:-Respbnse Rielationshap. ~ Q 1V ~ A phenomenological' exposure-response relationship n is now derived based on consistency (Hirayama, 19831b) Go ~ ~ A1#-6

I' (:anorr rtsk irom passi.e smoktng, 9' 3 I of' evidence provided by studies of lung cancer in non- smokers and from our exposure assessment. The Sevenrh . Day Adventist Study by Phillips er al: (1'980a, 1980b) appears to provide the best evidence of the magnitude of the lung cancer effect from passive smoking among U.S, nonsmokers. A calculation (Appendix C) based onithe age-stan- dardized differenres in, lung cancer mortality rates be- tween SDAs who never smoked and demographically comparable non-SDAs who never smoked'(age groups. 35' to 85+) from the studies of' Phillips et al: (1980a, 1'980b), yields an estimated 4700' lung, cancer, deaths. (LCDs) for the 62.4I million U:S'. nonsmokers (USDC. 1980) at risk (USSG~ 1979) aged Z, 35 yr. This in turn yields a passive smoking risk rate of' 7.4 LCDs per 100,000 person-yr (4700 LCDs/yr per 62,424',000 persons), inn good agreement with the value of'6.8 per 100,000 per- son-yr reported in the'Hirayamal(1981) study; To place the estimated mortality in perspectiwe„4700 deaths was about 9oV of the total annual LCDs, and'about 30?'0 of the LCDs in nonsmokers ini 1'9g2 (USSG, 1982): The exposure of'nonsmokers in the U.S. population of working age, taken from the model results in Table 1, appears to be a weighted average of'about 1.43 mg of tobaccotar per day, including the estimated I4e/o of the popuiationwho received no exposure at home or work. The carcinoBenic risks willlbe assumed to apply'even to retired persons, whose exp©sures are reported to be Iess than the'employed (Friedman er Ql.', 1983), because the risks of lung cancer from smoking decline only slowly even w,ith total cessation of exposure (USSG, 1982)„and because the risks of lung caneerincrease exponentially with age (,tiCI, 1966).. Using the statistical risk of 7.4'LCDs per 1100.000, andd dividing bythe'average exposure' of 1.43 mg/day, we estimate' a phenomenological exposure-response rela- tion appropriate for the general U.S. population at risk, of' about 5 LCDs per 100,000' person-yr at risk per, I mgi day nominal I exposure. The range'in1 nominal exposure has been estimated to be 0-14 mg/das• (ltepace and Lowrey, 1980). Studies of lung, cancer, and passive smoking across three cultures have shown an exposure-response relationship. Thus, the assumption of an exposure-response relationship is justified„an&a linear exposure-response function (Doll and Peto, 1'981: IRLG. 1'979;, U.S. EPA. 1979; Crump et al:, 1976) is assumed'. With zero excess risk from tobacco: smoke for zero exposure, and apply;ing, the e'sposure-respons'e relationship derived above, with the maximum exposure of 14 mg/day, a' maximum risk of about (14' x 5) = 70 LCDs per 100,000 person-yr is estimated for the most-expose&Iifestyle. This lifestyle has been previously typified by that of a nonsmoking musiciamwho performs regularly in a smoky nightclub and .. ho resides in a small apartment: w,inh a chainsmoker; many other scenarios may be draxn Low rey, 1980). t Repace and Estimated Loss of Life Expectancy R'rif (1981a, 1'981'b)i argues *that there exists a genetically determined distribution in natural' suscep» tibility, to lung cancer in people; the effect of exposure to tobacco smoke is to shift this distribution toward death at earlier ages. In other words; exposure to tobacco smoke produces a loss of life expectancy. One method of presenting risk data irtvolves calculation of the los's of Gfe expectancy, in units of days of'life lbst per inditi7dual, averaged over the entire population at risk. When ~ the average life-loss is muitiplied by the number of indi- viduals at risk, the impact: of the hazard on' society, in person-yr of life lost can be assessed. More itnportant4 one can display the age-specific probabilities of' death from the hazard, as well!as the average number of'years of life'lost by the average victim. Appendix C gives thee method of'calculation. Averaged over, a'Il! of the population at risk, (i.e., in- cluding those who die of other causes), the average l'oss of life expectancy from passive smoking is calculated' (appendix C) to be LS~ days, which is equivalent to an ultimate loss ofi2.S million person-yr ofilife forthrtotal at-risk U.S: population in 1979 over 33' yr of age'(62.4 million persons). The estimated worst-case' loss of life expectancy is 148 days, again averaged over all of the population at risk. The estimated mean Gfe expectancy lost by a passive-smoking lung cancer vicrim is ~ 17 * 9'yr. How does thccalculated average loss of life expectancyy for heavy passive smoking compare with the average loss of life expectancy found in active smokers? The modeled worst.case lifestyle might be reasonably ex- pected to have lesser exposure, and hence lesser risk than active smokers. Table 3, adapted from Cohen and Lee (1979)1 gives this comparison. The estimated most- exposed lifestyle has about I/j the loss of life expectancy of' the average pipe smoker, about t/• the lrass of the Tiable3. Estimated',loss of life expectancy'from aoti.^esmoking (all causes) and passive smoking (tung,cancer only),, adapted' from Cohen and Lee (1979Y. Cause Dars Cigarette, smoktng -malc 2250 Cigarette smok'tng-female BOfJ Cigar smoking, 330 Pipe smoking 220 N Passive Smok~inga (est. most~exposed tifestyle) 148 ' D Passivrsmokmg! (est. average lifestyle) 16' T,1. +Esttmted nhis work(see Appendix C);;averagedlover all nonsmokersA at risk. i.e., those N ho are presumed to die from passive smokangc~ induced lung cancer, and' those who do not. Estimates giM1en forGo passi.e smoking are phenomenologtcal estimates. rA X AA-7'

to, average cigar smoker, attd''/,o of that for active cigarettee smoking. Estimate of an Exposure-Rlesponse Relationship. Based on Ri'sks in Smokers An alternative extrapolated' exposure•response rela- tionship is now, dderived from evid'ence provided by studies of ' lung, cancer in cigarette smokers. Using the Surgeon Generalis estimate that 85 Qld of all'.lung cancers are due to smoking (USSG, 1982){ a: current annual. LCD rate to smokers at risk of about 3'1i6'per 100,000 is estimated (see Appendix B). Assuming a one-hit model (see Appendix B) for extrapolation of the risk (which in this range is functionally equivalent to the linear assump- tion that a milligram of tobacco tar inhaled by a non- smoker produces a response equivalent to that in a smoker) yields anestintate of'about0:87'LCDsI'1I00,00i01 person-yri. This corresponds to an exposure-response relationship of'0:6 LCDs/ 100,000 person-yr per mg/ day, and an annual aggregate risk estimate of about 555 LCDs per year, an, order, of'magnitude lower than the phenomenological, estimate. Discussion of Alternative Exposure-Response Relationships We now speculate on why these two different methods produce such disparate estimates ofl nsk. One possibility is that nonsmokers may have a reduced tol- erance to the effects of tobacco smoke. Another possi= bility is a"large dose" effect (Jarvis and Russell, in press), whereby exposure to tobacco tar at the lesser dlases experiencedi by nonsmokers produees, a greater risk per unicdose than the greater doses experienced by active smokers,, whose lung tissuri's saturated by car- cinogenic tar,. Large dose effects have been observed in cancer induction by ionizing radiation, in which the dbse-response curve has a linear form at low doses, a quadratic upwardl (positive) curvature at intermediate doses, but a, downward (negative) curvature at high doses (NRC, 198p),. Downturns in exposure-response curces of lung cancer, in smokers of more than 40:ciga- rettes per day have been observed by Doll and Peto (19'8),amd Hlirayama (1974). The effect of a leveling-off or dos.nturn ini the exposure-response curve at large ex- posures would be to cause ai linear model to underesti- mate the risk when extrapolated (i}Hoel et al., 1975', 1983: NRC, l98®)lover two orders ofmagnitude to low exposures. A third possibilit}•, is generated byy modeling the dose, as opposed to the exposure, of nonsmokers to tobacco: smoke. tionsmokers"exposure is translated into dose by means of a simple single-compartment model for lung deposition; and clearance (Repace. 1983') This modell suggetits that tar may accumulate on the surface of nonsmokers' lungs to an equilibrium, dose an ord'er of J. L. tttpaoe and A. H: Lo.Yra>y magnitude higher than the nominaliexposure, to1 a level of about 16 mg/day, due to the long pulmonary resi- dence times for respirable aerosols. If this 16mg,dose, rather than the 1.4-mg exposure; is the operative factor, then the typical' passive smoker would' have a risk, ac- cording to throne-hit model, ofiabout 9 per 100;000, in agreement with the phenotnenological estimate. There is support for this argument from, Matsukura% study (1984), which showed that heavy passivrsmokers had urinary cotinine levels comparable to active smokers of less than 3 cigarettes per day, and from Kasvga'S study (1983); which also showed that heavy passive smokers had urinary hydroxyproline levels almost equi- valent to that of light smokers. Moreover, similar ob+ servations have been found indicating that serum thio. cyanate (Cohen andl Bartsch, 1980) and benzpyrene (Itepetto and Martinez,: 1974) levels in some passive smokers were comparable to the elevated levels typicallyy found in smokers. The simple model'we have proposed ignores the effectt of cancer latency. The long latency period for lung cancer indicates that childhood passive smoking may be animportant factor affecting risk in adult life: Doll and' Peto (1981) have suggested that the effect of passive smoking may be surprisingly large because lifelong ex- posure mayy produce a lung-cancer effect four times as great as that which is limited to adult life (recalll the observation of Sandier et'ol:, in press: ehildhood'passive smoking appearedito elevate the cancer risk of adults). As Bonham and Wilson (1981) have shown from a na- tional study of 40,000children!in 1970, 62% came from, homeswith one or, morrsmokers, indicating that many adults receive exposure during childhood. Sensitivity Analysis Whii:h of the two exposure-response relationships derived is more useful in explaining actual epidemio- logical data? The Garfnnkel ('1981), American Cancer Society (ACS)! studpy ofpassivrsmoking and lung cancer, which spanned the years 1960 to 1972, reported' a stan- dardized mortality ratio of 1.20 and an annuall lung cancer rate of! 13.3 per 100;000 person-yr. Of the 176,739 women in the Garfinkel study, 28i°%o hadg non- smoking husbands. Thus, the "controls"' numbered 49.487 andlthe total "exposed"'were 127.252. According to census data ('U:S. Dept. of Commerce. 1980J: female participationirates in the labor force ra!ngedlfrom 37.1Q'o in 1960 to 3'8,8e'o in 1965', 42.8'% in 1970, and 43.707o in 1975, and were about 80o1a of the 1'965 lesel in 1947. Thus, it appears t~hat, about 380`0 of the women in this stud> were in the labor fr7rce: and presumably exposed to passive smoking,while at work. It is assumed that for both groups of women, control and exposed. 3'8°'o Kere employed and exposed to amibienotobacco smoke while at work. A_N indicatiedl in Table 11, typical U.S. nonsmoking AA-B

rancerrisk ftom passiNrsmoktng, Table 4, Number of women in each exposure category in the Garfinkel', (1!98i11 study of passive smoking and lung cancer. Group Number ~ Total cohon, 176;739 "True" controls; do nor,work,busbands do not smoke 30.682' 'Tainted"controls: wornk, husbands do not smoke 18;805 Total "eont rois"' 49,487 "Exposed- workers work.,husbands smoke 48:356' 'Exposed" nonMorkers: do not work„tiusbands smoke. 78.896 Tota! I'errposcdr 127y252 adults are estimated to inhale 1.82'mg of tobacco tarr per average day at work and 0:45'rng per average day at home, an exposure ratio of 4:1. This occurs becail although domestic and'workplace air exchange rates are sirnilar, ((Appendix A),,,workplace smoker densities tendi to be. far higher. Let the assumed basal rate of lung cancer deaths in these women from causes other than passive smoking be 8.7 ' per, 1i00,0W (the age-adjusted rate for nonsmoking women married to nonsmokers in Hirayarrta's study; P981Pa),. The Garfinkel (1981) ACS cohort can now be broken down las shown in Table 4. The Garfinkel (1981) study ean! be analiyzedl as follows, using the phenomenological exposure-response relationship of 5 LCDs/'1I00,00p person-yr-mg/day. The lung cancer deaths per, 100,000 contributed byy passive smoking,are then 2.25' (0:45 x 5) for the home andl9!10 (1.82 x 5)! fior the workplace. Application of these figures to the numbers of true and'tainted controls and! working and nonworking, exposed women yields, after addition of the basal risk of 8.7 per 1:00,000, the estimated rates for lung cancer deaths.per 100,000 per- son-sr, as shown in Tablt 5. The ratio of risks (all ex- posed:all controls) is thus 1.19. The ratio (averaged over husbands' heavy and light smoking categories) in the Garfinkel1 (1981) study was 1.20. Itss than, a 1I dif- ference. The lung cancer death rate for the weighted average of the "exposed" and' "control" categories is 13.8 per 100,000. Over the 12'yr of the Garfinkel study, the actuallrate averaged 13:3 per 100,000, a less than 40,10 difference. In ooher words, this analysis (Repace, 1994) appears to explain botih the observed lung cancer death rate and obsen•edlrisk-ratio of'the GarfinkellACS cohl Could this be due to chance?'Suppose that, instead of 3!8PO of wmmen in the workflorce, 1001ro of women Tabic 5 C.ai.ulaieCllun€ cancer nsksfor each subgroup in the Gar;ink'.ef 119,€11 studs using,the 5 LCUa 100.000 0 prssun-cr mg dax erposure•response relatton. Group Rate True cuntrols Tiarnted controls I 7.8' u8 '• 9.1 Y Alllcomrob (MerQhted mean) 1:L16 Esno,cd workris 20.tl5/x' . _._` - 9 101l Erpo.sed nonwo:rkenAL"e%posed,IMaghted mean> 10.9K Itl.' • 117 411 :..51 worked. Then the ratio of'risks would be 1.13, ,a 6%ldif- ference from, Garfinkel's observation, but the annual lung canceri death rate would be 19:42, a 46076 dif- ference: Suppose09lo of women worked. Then the ratio of risks would be 1.26, a 5e/o difference from Garfinkel's result, but the lung eancer, death rate would be 10.32 per 100,000, a 22'o1a difference from Garttnkel's observation. Suppose the exposure-response relationship of 0.6' LCDs/ 100,0'00~ person-yr per mgJday yielded'. by extra- polation with the one-hit model from ~the risks in smokers is used! The lung cancer deaths per 100,000 contributed by passive smoking are then 0.27' (0.45 x 0:6)! for the home and 1.1 (1.82' x 0.6) forthrworkplace. Applica, tion of'these figures to the numbers of true and tainted controls and working and nonworking exposed women yields, after addition of the basal risk of 8.7 per 10Q,!000, thrfigures shown in Table 6. The ratio of risks: (all exposedtall controls) is then.1.03'. Compared withl the risk ratio in the Garfinkel (1981) study, this is a 1ll difference. The lung cancer death rate forthe weighted, average of the "exposed" and "control" categories is 9.3' per 100;.000, a 30"lo difference froml Garfinkel's result. When the one-hit model is used, the ratio of "al1- exposed"' to "true"' controls l is 1.09, a: 38076 di'fferencre with Hirayarna's ratio. The corresponding lung cancer mortality ratris 9.45, a 39% difference with Hirayama's result. Finally, using the phenomenological exposuraresportse relation„the ratio for "all exposed" and "true"'controls is 1.7. Hirayama's. (1!98'1') average risk ratio was 1i.78 from passive smoking, a 4.5101o difference. Furthermore, if lung cancer risk rate calculation is performed with the tainted'controls included'as an exposedlgroup; the result is 14.8 per 100,000, cbmpared with Hirayama's ob- served 1515~ per 1i00,000.; a 4% difference. In other words, the effect of moving, the confounding tainted controls from Garfinkel's control group into his ex- posed'.group is to yield results withinl51°'0 of Hirayama's. Thus, on the basis of'this sensitivity analysis, it would appear that the phenomenological exppsvre-response relationship is, better able to describe the results of the Garfinkef ('198'l ) stud~ tfian; the one-hit model., and in addition, also appears to be able to explain quantita- tively wh5• the two large prospective studies of passiva smoking and lung cancer yielded diffrrent, results. Tahle:6'. Calculatrd lung caneer risks for each subgroup in thr Garfinkel (1981) stud>,ustng,the 0.6LCOB per 1001WC1 prrson,+r mg day ecpvsure-response retauon. 6ruup Rate Taur controls 8.1 Trtmcd controls 9:8 t8!7 + 1.11 All controls pNetghtied meanl 9:11 Esqxs,cd "arkers l0!0" IB!~' + 0:=' . I .11 Eopu.cd nunMurkers 8.97 l8'J' + 0:."/' All "posed nwrrghted mean) 9;39 AA-9

0 12' Table 7' Comparison of estimated risks from various hazardous air, pollutants. Risks have been assessed for non-oecupational!exposures of the general population to several,hazardous air pollutants. Alllare airborne carcinogens; all but passive smoking are being regulated'by society. The'.statistical mortality gi.en is before control. Pollutant Estimated Annual Mortality+•e Reference Paasive,smoking, 5000 LCDs per yr (this work) Vinyl chloride <27 CDs per yr, (UiS. EPA, 1975): Radionuclides (worldwide impact from, Department of EnerBa facilities) Coke oven emissions Benzene Arsenic 7 CDs per yr <15 LCDs per yr <8CDs per yr, <S ~ LCDs per yr U.S. EPA. 1983b1. (UIS. EPA, 1984)', (UiS. EPA. 1979b). (UiS. EPA. 1980) aCD: _::ncer death: LCD s lung eancer.death: eRisks for passive smoking,and radionuclides are best estimates„and risks for other pollutants are upper bound. Comparison of the Estimated Risk of Passive Smoking with those of H'azardo'us'Air Pollurtants Cwrrently Under Regulation Although the.quantitatiive estimates presentedlshould be regarded as preliminary and subject to confirmation by further research, the evidence suggests that passive smoking appears to be. responsible for aboun one-third of the annual lung cancer mortality among U.S'. non, smokers. To place these estimates in perspective, Table 7' gives a comparison of the estimated risk, of passive smoking to risks estimated by'the U.S. Environmental Protection Agency for the carcinogenic hazardous air pollutants currently regulated under section, 112 of the Clean Air Act,(SCEP; 1977): As Table 7'demonstrates, passive smoking appears to pose a public health risk larBer than, the hazardous air pollutants from all regu- lated industrial'emissions cotnbined: Acknmwkdgrmenrs-7lhe authors are grateful to R. L. Phillips of the Department of Biostatisti¢s and Epidemtology of Lotna.Linda Uni.ror• sny.loma Linda. CA„ for tabulations from his published studies of mortality in members of the Seventh Day Adventist Church. Wralto thank B. Fischoff. H. Gibb. J. Horowitz. D. Patnick, G. Sugiyama, W. ©tt, and 1. We11S for useful discussions, and Jl DeVtoeker for assistance with computer programming. Appendix A: Modeling Exposure of, Nonsmoking U,.S, Adkulrts to Ambient Tobacco Smoke lntroductvon Lifestyle is the integratied'way oflif'e of anlindi's•idual; aspecrsof lifestyle which willbe considered here have,tio do with the amount of time a nonsmoker spends in con- tact with smokers, and therefbre with their effluent. Ex- posure of nonsmokers to tobacco smoke might be ex- pectied'to be commoni in the United States because one out of three U.S. adults smokes cigarettes at the J! L. Rtpace and A. H. ,,o„1eS• estimated rateof '3'2 per day(Rkpace and Lowrey, 1'980), and an additionallone ounof six smokes cigars or pipes. Furthermore., indoor'air, pollution from tobacco smoke' persists in indoor, environments long, after smoking, ceases (Repace and Lowrey, P980„ 1982),. Earlier work (Repace and Lowrey, 1980) presented aa model of nonsmokers' exposure to the particulate phase' of ambient smoke which was supported by controlled experiments' andl field survey of the'.levels of respirable particles indoors and out„in both smokefree and smoky, environments. This work, which established': that am- bient tobacco sm oke imposed' significant ai'r' pollution i burdens on nonsmokers, was extended by later work (Re- pace and Lowrey, 1982); that further demonstrated thee predictive power of this model. The. model predicts, a, range of exposure of from 0to 14 mg,oficigaretrcraerosol per day, depending upon the.nonsmoker"s lifestyle. Ex- posures of, prototypical'nonsmokers were modeled, bur' no attempt'was made to estitnate the average population exposure. Concentrations of ambient, tobacco smoke encountered. by nonsmokers can be approximated by equilibrium values that' are determined by the ratio of the average smoker density to the effective ventilation rate (R'epace and Lowrey, 1!980, 1982); in practice, db- sigtt ventilation standards': based on occupancy are use- ful surrogates for effective ventilation rates. On the average; a characteristic value of this ratio can be assigned to a particular microenvironmental'class, e:g., homes,, offices; restaurants, etc. (Repace er cf., 1!981D). Therefore, the.average daily exposure of'india,iduals can be estimated from the time-weighted sum of concentra- tions encountered, in various microenkironments eon- taining smoke (Ott,,in press; NRC, 1!98'L; Szalai, 1972; Repace et cl., 1980). Frposarre and lifestyle. It is important to realize that most' persons"lifestyles are such that they spend nearly 901r0 of their time.in just two microenvironmental classes, thus affording a great simplification of exposure. modeling. Szalai ('1972), as part of The Multinational Comparative Time Budget Research Project, which studied the, habits of nearly, 30,000: persons in 12 countries ('from 1964 t o 1 1966); h'as compiltrd' data reporting the average time spent in various locations or microenti•ironments. The'a data for. 44 cities in the United States, as,analyzed by Ott (in press) are summarized in, Table Al (',see also ItiiRC. 1981)., Table Al shows that U.S. urban; people spetid an average of 8'8o'o; of their, time in jisst two microenviron- ments: in homes and workplaces. Moreover, employed persons in the U.S. cities are estimated to spend only , 30'o of the day outdoors, while housewives. silend only 2'''o outdoors (Otn, in press;. NRC, 1981'). Assume that these values are representative of'the entire'populatibn. [In 1970, approximately three'fourths o'flthe population, was urban (USDC„ 1980),] AA-1 0