Tobacco Documents Online

Environme.u lnuernational. Vol. 14. pp. 249-265. 1988 Printed in the USA. All rights reserved. AN ESTIMATE OF ADULT MORTALITY IN THE UNITED STATES FROM PASSIVE SMOKING A. Judson Wells 102 Kildonan Glen. Wilmington, Delaware 19807. USA (Received 9 December 1987; Accepted 7 July 1988) Ij '~~ , 1S J1 uIGU.Jl2o.'xR S3.twl , .(K) Copyright .C 1988 Pcrpmon Press plc NOTIt;E This matotial msy be Votected by CWl& i4w (iitte 17 U.S. We). The purpose of this paper is to estimate the number of adult deaths per year in the United States from passive smoking. The epidemiological literature on passive smoking and adult mortality and cancer and hean morbidity is reviewed. Combined relative risks (or lung cancer, cancers other than lung. and heart disease are calculated for each sex and disease category. These data along with estimates of nonsmoker death rates and populations exposed allow calculation of annual deaths in each ca4egory. Reduced relative risk and reduced exposure at older ages are taken into account as well as a correction for possible misclassification of smokers as nonsmokers and exposed nonsmokers as nonexposed. Al- together 46.000 deaths per year are calculated consisting of lung cancer (3000) other cancer (11.000) and hean disease (32.000). Reasons why such high estimates for other cancer and hean disease may be possible are explored. It is concluded that exposure to environmental tobacco smoke can have adverse long term health effects that are more serious than previously thought. Introduction Several attempts have been made to estimate U.S. adult mortality from passive smoking. For example. Repace and Lowrey (1985) estimated the lung cancer deaths to be about 5000 per year. Fong (1982) estimated total mortality at 10.000 to 50,000. Russell et al. (1986) es- timated total U.S. mortality at more than 4000. The present estimate is based on epidemiological evidence currently available on lung cancer, cancers other than lung. and heart disease. The Surgeon General of the United States (USSG. 1986) and the U.S. National Academy of Sciences (NRC, 1986) have issued reports stating that passive smoking can cause lung cancer. In the National Acad- emy report the relative risks from the various lung can- cer studies were combined into an overall relative risk using a procedure somewhat similar to that which is used in this work. The Academy report then projects that about 20% of the 12,000 U.S. lung cancer deaths per year among never smokers is due to passive smok- ing. This is reasonably close to the 3000 per year pro- jecte.d here for never smokers plus exsmokers. The . methods used in the National Academy report are fur- ther detailed in Wald et a!. (1986). Blot and Fraumeni (1986) have also presented an overview of studies of lung cancer and passive smoking. They use a method of combining the relative risks from various studies es- sentially identical to that used here. Thus, the proce- dure of combining relative risks from various passive smoking studies to obtain overall relative risks and tighter confidence'intervals is now well established by authorities in the field. Also, the method used here to calculate annual deaths from the relative risks appears to be validated by the National Academy results for lung cancer. However, both the Surgeon General's task force and that of the National Academy felt that the data, as of 1986, on cancers other than lung and on heart disease were still too meager to allow calculation of reliable overall risks. Since 1985 considerably new epidemiological infor- mation has become available, particularly on heart dis- ease. This new information is reviewed and combined with the old data to calculate updated relative risks, overall confidence "limits, and estimated annual U.S. deaths from passive smoking and the three main dis- eases, namely, lung cancer, cancers other than lung, and ischemic heart~yI~ar' The total particulate matter i" dose retained by p 3~ve smokers is too low to account for the health eKe; s ol Oassive smoking if one starts with the health effe s exhibited by direct smokers and ratios down from,t e dose retained by them. Reasons why such a discrep ncy might occur are explored. 249

,50 r , A. J. wells . Methods - Studies to be considered in the analyses were ob- tained originally from the literature searches of the U.S. Office on Smoking and Health (OSH, 1979-85). More recently. studies have come to light primarily through personal contact with workers in the passive smoking field. Criteria for admitti~"g data to the analysis are: 1. Studies on the association of passive smoking with adult mortality or morbidity from lung cancer, other cancer or ischemic heart disease were included. All cause data were not used because essentially no male data are available. The female data, if calculated, yield overall results that are in the same range as the results derived from the three main diseases (see Appendix B). Emphysema is not included because the nonsmoker death rate is so low that less than 1% of deaths from passive smoking would be pre- dicted from this source (see Appendix B). 2. Retrospective studies should have controls. 3. Observations should be based on spouse exposure or on general exposure of more than 10 years du- ration. The diseases under study are known to have long induction periods, and it is assumed that most married people old enough to die of passive smoking would have been exposed 20 years or more. 4. Enough data should be available from the study to allow calculation of a weighting factor for combining the relative risks. Two risk models were used and a third was consid- ered. The primary model used combined relative risks from the various studies that pertained to a given sex and disease and assumed that the combined relative risk was constant with age, although variation with age of the underlying neversmoker death rate and the fraction of the population exposed were included. In the sec- ondary risk model the combined relative risk was also allowed to vary with age. These models were suggested in part by the considerations in James Robins' Appen- dix D in the National Academy report (NRC, 1986). The third risk model was based on the rate difference between the death rates for exposed and nonexposed populations. A detailed analysis of this model for heart disease in women was carried out (see Appendix C). It was concluded that the relative risk models were much superior to the rate difference model when combining data across different cultures as is the case here where some of the studies are from the orient. Wherever a study showed both a crude relative risk or odds ratio and an adjusted ratio, the adjusted ratio was used. To obtain a combined relative risk a method similar to that of Blot and Fraumeni (1986) was used. Case control studies were aggregated using Program 2 of Rothman and Boice (1982). Cohort studies were ag- gregated using Program 7. A combined relative risk for the two aggregates was obtained using: Rte s expw`°InR,, + w«InRn (1) K'co + K'cc where &; Rro, and RK are the relative risks for the combined total, the cohort studies, and the case control studies, respectively, and w,, and w« are the weights for the cohort aind case control studies. respectively. which are the inverse of the respective variances. Vari- ance is taken as the square of the standard deviation which is equal to In R/x, so the weight, w=(X/In R)'-. The source of these equations is Rothman (1986). Con- fidence intervals were calculated from a combined X= w'"- In RN.'fior some studies it was necessary to calculate a chi from the confidence limits in order to calculate a weight since no other data were available. These data were then combined with the rest using Eq. (1). Ages of death from 35 and up were used and should include essentially all adult deaths from passive smoking. In some studits morbidity relative risks were reported whereas our interest is in mortality. The morbidity rel- ative risks were accepted as surrogates for mortality relative risks because, for cancer, the survival rates for exposed and nonexposed cases appeared to be similar while, for heart disease, incidence relative risks, if any- thing, are lower than mortality relative risks (Svendsen et al., 1987). The 1985 smoking status for U.S. residents in 5 year age increments was obtained from the National Center for Health Statistics. Nonsmokers were equated to never smokers plus exsmokers. The fractions of never smokers living with ever smokers (24% for mates and 60% for females), all of whom were considered to be exposed, were obtained from controls of the U.S. based studies for all three diseases. These fractions were as- sumed to hold also for nonsmokers (never plus ex). The fractions of all nonsmokers exposed as nonsmokers liv- ing with nonsmokers, but still exposed at home or at work (37% for males and 16% for females), were ob- tained from Friedman val. (1983). These fractions were assumed to hold for nonsmokers living with never smok- ers. By adding the two fractions the total nonsmoker exposure of 61% for males and 76% for females was obtained. These overall exposure fractions are known to be higher at younger ages and lower at older ages. The data of Friedman etal. (1983) were used to develop smoothed values of fraction exposed 10 years earlier (midpoint of `a 20 year exposure) for each sex and 5 year age interval normalized to 61% for males and 76% for females. by multiplying each population element by.each fraction exposed element, the exposed popu- lation by sex and 5 year age interval could be deter- mined. Death rates for never smokers for lung cancer by sex and 5 year intervals were drawn from Garfinkel (1981) 50986 1060

. I , I Adult mortality from passivc smoking and smoothed using a semi-log plot against age. For cancers other than lung-for females a semi-log plot of 1984 age specific death rates for ages 35 + was devel- oped for'malignant neoplasms less malignant respira- tory neoplasms from the data of the National Center for Health Statistics (1986). Then, a parallel plot was developed using as reference points the neversmoker data of Hammond (1966) for ages 45-64 and 65-79 to yield neversmoker rates for ages 35+ for each 5 year age interval. For heart disease never smoker death rates by sex and 5 year age intervals for 1963 were developed from the appendix tables in Hammond (1966). These were reduced to 1984 equivalent iates (with the reduc- tion factors corrected for the effects of smoking) by a technique similar to that used by the U.S. Office of Technology Assessment (OTA, 1985). Semi-log graphs were used to estimate never smoker death rates by 5 year age intervals for the entire age range (see Appen- dix A. Table A3). The excess death rate for never smokers for passive smoking (D,,,) for each sex. disease and 5 year age range was calculated from the never smoker death rates (D,v) using the formula: Do, = D,JR - 1)l(Fo(R - 1) + 1) (2) where Fo is the fraction of the population that is exposed and R is the combined relative risk. This excess death rate was assumed to apply to all nonsmokers. Deaths were then calculated by multiplying the passive smoking excess death rate by the exposed population for each sex and 5 year age interval, and summed. For those calculations where the relative risk was assumed to have varied with age, the excess death rates for passive smok- ing were recalculated from the age specific relative risks for each 5 year age interval. Additional calculations were carried out to show the effects of bias including those from misclassification of smokers as nonsmokers and exposed nonsmokers as unexposed, using a method similar to that of Wald et at. (1986). Results Relative risks The results for passive smoking relative risk for fe- males for lung cancer are shown in Table 1. The three cohort studies are listed first and show a combined rel- ative risk for all exposures including exposures to exsmokers of 1.34. At the time the analysts was made there were fourteen acceptable case control studies with a combined relative risk of 1.50. The overall combined relative risk, based on 1,174 cases, is 1.44 with 95% confidence limits of 1.3-1.7. The male lung cancer ob- served relative risks are shown in Table 2. There are now nine studies with 144 total cases. The overall com- bined relative risk is 2.1 with 95% confidence limits of 251 1.3-3.2. Data excluded from Tables'l and 2 along with the reasons were the following: Chan et at (1979), cur- rent exposure only; Knoth cr al. (1983), no controls; Kabat and Wynder (1984) nonspouse data, current ex- posure only; Buffier et a1. (1984), 0-32 year data, not a minimum bf 10 years exposure. A paper by Dalager et al. (1986) describes a pooling of data from Correa et at (1983), B„uffler et at (1984) and a study of males in New Jersey. ; I'hey observed an adjusted odds ratio for spouse exposure of 1.47, but since Correa et at (1983), and Buffler tt at (1984), were already included in Ta- bles I and 2 and since the New Jersey data were not available separately, it was decided to omit the Dalager et at. (1986),study from this analysis. Also. available were abstracts of two recent papers. Geng et at. (1987) from China t+vith a relative risk of 2.2 and Inoue and Hirayama (1087) from Japan with a relative risk of 2.3, both for females. Also W. K. Lam (1985) in a thesis from the University of Hong Kong that is quoted in Lam et al. (1987) found a relative risk of 2.0 for ad- enocarcinoma among females. These inputs arrived too late to be included in the analysis. . The data of Hirayama (1984a) on female lung cancer are sufficiently detailed to indicate a declining relative risk with age from 1.87 at approximately age 50 to 1.43 at approximately age 75. These data were used to de- velop a second death calculation assuming a declining relative risk, but still normalized to 1.44. However, Hirayama's data show no such decline in passive smok- ing relative risk with age for males. Instead, the trend appears to rise with age, so no secondary calculation was made. There are now five studies relating passive smoking to total cancer or cancer other than lung in females. The individual and combined relative risks for females are shown in Table 3. The total combined relative risk is 1.16. The total cases, 2,933, are two and one-half times the total cases for female lung cancer (Table 1) although 2,505 are concentrated in the large Hirayama (1984a) study.'This is a large data base. The total com- bined chi square is 11 compared to 27 for female lung cancer. The two largest of the female studies. Hirayama (1984a) and Sandler et a!. (1985), cover different age of death ranges. Hirayama covers 50 to 80+ while Sandler et al. cover <30 to 59. The two studies taken together would indicate a rather sharp decline in rela- tive risk with age from about 3.5 at age 40 to about 1.04 at age 80. The' high relative risks at the younger ages may be due to premenopausal breast cancer (see San- dler et al., 1986). Two calculations of U.S. female deaths from passive smoking and other cancers were made, one using the 1.16 relative risk from Table 3 at all ages and one using the declining relative risks. Gillis et al. (1984). Sandler et al. (1985). and Rey- nolds (private communication) also report on other can- 50986 1061

•252 Table 1. Female relative risks for lung cancer from passive smoking. A. J. Wells . T Highest Exposure All Exposures Mantel Trend Locale otal Cases RR 2-tail p RR 95 % C.L. 1-tail p Cohort Studies: Hirayama (1984a) Ja¢5n 200 1.9 0.002 1.6 1.1- 2. 2 0.002 Garfinkel (1981) United States 153 1.1 - 1.2 0.8-1.6 - Gillis er a1.'(1984) Scotland 8 1.1 0.2-5.6 Combined Cohort 361 1.34 1.1-1.7 Case Control Studies: Trichopoulos et at. (1983) Greece 77 2.6 0.19 2.1 1.2-3.6 0.005 Correa et al. (1983) Louisiana 22 3.5 0.02 2.1 0.8-5.2 Buffler et at. (19&1) Texas 27' - - 0.9 0.4-2.3 Kabat and Wynder (1984) United States 24 0.8 0.3-2.5 Sandler et al. (1985) Garfinkel et al. (1985) North Carolina United States 2 116 2.0 0.05 inf 1.3 0.8-1.9 0.025 Wu er al. (1985) California 28' 1.2 0.5-3.3 Lee et al. (1986) United Kingdom 32 - - , 1.0 0.4-2.7 Akiba et al. (1986) Japan 94 2.1 - 1.5 0.9-2.6 0.06 Koo ct at. (1987) Hong Kong 86 1.2 - 1.6 0.9-3.1 Pershagen et a!. (1987) Sweden 67 3.2 - 1.2 0.7-2.1 0.12 Humble et al. (1987) New Mexico 20 1.2 -- 2.3 0.9-6.6 Brownson et al. (1987) Colorado 19 - - 1.7 0.4-3.0 Lam et al. (1987) Hong Kong 199 1.65 1.2-2.4 Combined Case Control 813 1.50 1.3-1.8 Combined Cohort and C/C 1174 1.44 1.26-1.66 ' Private communication. 'From Blot and Fraumeni (1986). cer in males. The relative risks were 0.6. 1.5 and near unity, respectively. The number of cases in each study is very small with no statistical significance. Therefore, it was decided to use a neutral relative risk of 1.0 for males for cancer other than lung until more data become available. There are now six studies of passive smoking and heart disease in females. The individual and combined relative risks are shown in Table 4. Studies new since 1985 are Lee et dl. (1986), Martin et al. (1986a) and the important, largeHklsing et al. (1988) paper from Mary- land. The overall combined relative risk based on 1.622 cases is 1.23 with 95% confidence limits of 1.11 to 1.36 and a combined phi square of 16. Helsing et al. (1988) and Martin et al. (1986a) provide data for younger women and indicate high relative risks (average 2.45) Table 2. Male relative risks for lung cancer from passive smoking. Locale Total Cases RR Cohort Studies Hirayama (1984a) Japan 64 2.3 Gillis et at (1984) Scotland 6 Combined Cohort 70 Case Control Studies: Correa et at (1983) Louisiana 8 Buffler et at (1984) Texas 8' Kabat and Wynder (1984) United States 12 Lee et at. (1986) United Kingdom 1S Akiba et at (1986) Japan 19 Humble et at (1987)' New Mexico 8 Brownson et at (1987)' Colorado 4. Combined Case Control 74 Combined Cohort and C/C 144 'Private Communication. Highest Exposure 2-tail p, 0.16 All Exposures Mantel Trend RR 95'k C.L. I-tail p 2.25 1.11- 4.9 0.023 3.3 0.7 -16.5 2.5 1.2 - 5.0 2.0 0.4 -10 1.6 0.3 - 8.1 1.0 0.3 - 3.2 1.3 0.4 - 4.6 1.8 0.5 - 5.6 4.2 1.0 -16.8 2.7 0.2 -31 1.$ 1.0 - 3.3 2.1 1.3 - 3.2

, Adult mortality from passive smoking 253 . Table 3. Female relative risks for cancer other than lung"from passive smoking. Highest All Mantel Exposure Exposures Trend Locale Total Cases RR 2-tall p RR 95 % C.L. 1-tail p ~..._. • Cohort Studies: Hirayama (1984a)' y. Japan 2505 1.16 0.01 1.11 1.0 -1.2 0.05 Gillis et al. (1984) Scotland 43 - - 1.2 0.6 -2.5 - Reynolds ei al. (1987) California 70° 1.7 1.1 -2.7 I Combined Cohort 2618 1.13 1.03-1.24 Case Control Studies: Miller (1984Y Pennsylvania 84 1.25 0.7 -2.3 Sandier et a!. (1985) North Carolina 231 2.0 1.3 -2.9 Combined Case Control 315 1.7 1.2 -2.45 Combined Cohort and 2933 1.16 1.06-1.27 C!C I t 'Obtained by subtracting data for lung cancer from data for all sites. 'Provided by Dr. Reynolds. 'Age adjusted Mantel-Haenszel values for nonemployed wives. for ages up to about 50. At higher ages there is no trend with an average relative risk of 1.17 holding out to age 84. For male heart disease and passive smoking there are now four studies (see Table 4). The two new ones are Lee trt al. (1986) and Helsing et al. (1988). The result of Svendsen et al. (1987) is shown for information, but is not used in calculating the combined relative risk because it pertains to a high risk group. The combined relative risk based on 443 cases is 1.31 with 95% con- fidence limits of 1.1 to 1.6 and a combined chi square of 9. The results are remarkably uniform. As in the female data the relative risk is high at the younger ages, about 2,.9, but declines to a nontrend average of 1.28 which extends from age 55 out to the older ages. Svendsen et al. (1987) show that there was very little difference between never smoking men married to nonsmokers and those married to smokers in the major coronary risk factors such as baseline blood pressure, total cholesterol, and LDL cholesterol. This work was reported in more detail in Martin et al. (1986b). Small differences were found in weight (195 vs. 190 if wives were smokers) and drinks per week (10 vs. 8 if wives were smokers). On the other hand, Garland et a!. (1985) Table 4. Relative risks for heart disease from passive smoking. Highest All Mantel Exposure Exposures Trend Locale Total Cases RR 2-tail p RR 95 % C.L. 1-tail p I t Females Cohort Studies: Hirayama (1984b) apan 94 .3 0.038 .16 .9- 1.4 .02 Gillis et a!. (1984) Scotland 21 3.6 0.9-13.8 Garland et al. (1985) California 19 - 3.5 0.9-13.6 Helsing et al. (1988) Maryland 988 1.27 - 1.24 1.1- 1.4 0.005• Combined Cohort 1522 i 1.23 1.1- 1.4 Case Control Studies: Lee et a!. (1986) United Kingdom 77 0.9 0.7- 1.3 Martin et a!. (1986a) Utah 23 2.6 1.2- 5.7 Combined Case Control 100 1.29 0.8- 2.0 Combined Cohon and C/C 1622 1.23 1.1- 1.4 Ma/es Cohort Studies: Gillis et a1. (1984) Scotland 32 1.30 0.7- 2.6 Lee et al. (1986) United Kingdom 41 1.24 0.5- 2.6 Helsing et a/. (1988) Maryland 370 1.31 1.1- 1.6 Combined Cohort 443 1.31 1.1- 1.6 Svendsen et al. (1987)° United States 13 2.2 0.7- 6.9 'Based on Cochran chi-square of 9.2. •MRFIT cohort of high risk individuals, included for information only. 50986 1063

. . found that never smoking women married to smokers had slightly lower weight, slighily lower blood pressure, and slightly higher cholesterol, all nonsignificantly dif- ferent, versus never smoking women married to never smokers. All of these authors conclude that the in- creased passive smoking risks they observed cannot be ascribed to differences in the m.pjor coronary risk fac- tors between passively exposed and nonexposed never smokers. It is impressive that the relative risks for heart disease from passive smoking rise in an orderly manner from the lowest risk group. Japanese women at 1.16, through American women at 1.27, and American men at 1.31, to high risk American men at 2.2. A correction for misclassification was attempted for all three disease categories. Following Wald et al. (1986), and presuming that the passive smoking studies were done somewhat more carefully than the general questionnaire studies they cite, it was assumed that 5% of ever smokers were misclassified as never smokers. Along with Wald et al. (1986) we assumed that the nonexposed nonsmokers were actually exposed to 1/3 the extent of the exposed nonsmokers except that for Greece. Japan, and Hong Kong, where less than 30% of women had ever smoked, the correction for nonex- posed female nonsmokers was omitted. It is believed that older. nonsmoking women in Greece and Japan, and probably in Hong Kong also. because of their social habits, were exposed to relatively little tobacco smoke beyond that of their husband's. Since most of the mis- classified smokers were found to be light smokers or longstanding exsmokers, reduced relative risks for the misclassified ever smokers were calculated, as noted in Appendix A. The modified passive smoking relative risks are shown in Table 5. The false relative risks due to smoker misclassification are somewhat lower than calculated earlier by Wells (1986) because of the as- sumption of light smokers and long term exsmokers among those misclassified. following Wald-eral. (1986), and the use of a more accurate formula. In general, the misclassification of smokers has a large negative effect on male relative risk which is more or less offset by the positive effect of exposure of the "nonexposed." For females the smoktr,misclassification effect is small fo negligible, but because the relative risks are smaller and no correction was made to "eastern" data (Japan. Greece, and Hong Kong), the positive effects of ex- posure of "nonexpo$ed" are also smaller. Calculation of Deaths The details for the calculation of female lung cancer deaths from the relative risks, both constant and de- clining, are shown°in Table 6 as an example. Similar" calculations were made for the other disease and sex categories and are 'shown in Appendix A. The results of all of the calculations are summarized in Table 7. These results are restated per million total population in Table 8. Where khe relative risk appears to decline with age and where neversmoker death rates at the younger ages are low, as in female heart disease and lung cancer, there is a reduction in mortality calculated by using the age specific relative risks. Otherwise, the higher exposed population at the younger ages out- weighs the higher death rate at older ages and total mortality is increased. In terms of total deaths the ef- fects of using age specific relative risks tend to cancel out. The total deaths, before adjustment for misclas- sification, for both males and females are about 19,500 for a total for both sexes of about 39,000. The effects of misclassification on total deaths are substantial, raising the total to 53,000. Most of this increase is in heart disease where the numbers are large and the effects of smoker misclassification. although not necessarily small, are still heavily outweighed by the partial exposure of the "nonexposed." To be conservative a best estimate for passive smok- Table 5. Passive smoking relative risks modified for misclassification. A. J. Wells Lung Cancer Other Cancer Heart Disease Females 1. Combined relative risk. 1.44 1.16 1.23 2. False relative risk due to projected 5% smoker misclassification. 1.01 1.002 1.01 3. Combined relative risk corrected for smoker misclassification, (1) + (2). 1.43 1.16 1.22 4. (3) corrected for exposure of "non• exposed" at 1/3 that of exposed. 1.48 1.21 1.32 Males 1. Combined relative risk. 2.1 1.01 1.31 2. False relative risk due to projected 5% smoker misclassification. 1.3 1.11 3. Combined relative risk corrected for smoker misclassification. (1) + (2). 1.6 1.17 4. (3) corrected for exposure of "non- exposed" at 1/3 that of exposed. 2.4 1.29 'Assumed value for lack of better data.

'• Adult mortality from passive smoking 255 • Table 6. Annual U. S. female lung cancer deaths from'passive smoking. d E Relative Risk Constant at 1.44 Relative Risk D li i Age of Death Neversmoker Death Rate per 100.000 f' Nonsmoker Population 1000's Fraction Exposed xpose Population 1000's Excess Death Rate Deaths RR ec n ng Deaths 35-39 1.6 6150 0.94 5781 0.50 29 1.70 39 i0-44 4 2 4622 0.92 4252 0.75 32 1.69 43 t 1 45-49 . 3.6 3846 0.89 3423 1.14 39 1.68 52 50-54 5.3 3856 0.87 3355 1.69 57 1.62 72 55-59 7.8 4161 0.84 3495 2.51 88 1.56 104 60-64 11.0 4192 0.77 3228 3.62 117 1.49 126 65-69 16.6 4160 0.70 2912 5.55 162 1.43 159 70-74 23.5 3441 0.59 2030 8.21 167 1.36 142 75-79 34 3004 0.49 1472 12.3 181 1.29 127 80-84 46 1886 0.29 547 18.0 98 1.18 43 85 + 52 1003 0.10 l00 21.9 22 1.0(t 4 Totals 13.0 40291 0.76 30595 3.0 992 - 911 ing deaths might be 46,000. half way between the 39.000 calculated directly from the relative risks and the 53.000 calculated using the modified relative risks. By disease the total would consist of 3,000 lung cancer. 11,000 other cancer, and 32,000 heart disease. For each million of total population the deaths by disease would be 13 for lung cancer. 46 for other cancers, and 134 for heart disease. These numbers may be useful for populations similar to that of the United States in terms of propor- tions of never smokers, exsmokers, and smokers, and in terms of the proportion of the population that is less than 35 relative to that over 35. For other populations the per million numbers are best not used, but the meth- odology can be used. That cancer other than lung and heart disease are legitimate contributors to deaths from passive smoking is supported in Hirayama. (1984a,b) in his large prospective study. He found significantly elevated risks for all three diseases. and his result for lung cancer is now believed to be valid. (USSO 1986; NRC, 1986). It is difficult to believe that his lung cancer result is valid while the other two are not. Discussion The cancer sites for passive smoking appear to differ somewhat from those for direct smoking. Using infor- mation on specific cancer sites from Dr. Hirayama (pri- vate communication) it appears that cancers common to both types bf smoking are lung, liver, cervix, nasal sinus, and leukemia. Some of these cancers are on1Y weakly associated with direct smoking. Cancers asso- ciated to som degree with direct smoking, but absent in passive sm~king are buccal cavity. pharynx, larynx. esophagus, stomach (Hirayama, 1984a), urinary blad- der (Kabat et 01., 1986), kidney and pancreas. Cancers related to passive smoking, but absent in direct smoking are brain (Hirayama.1984a), endocrine glands (Sandler et aL,1985), lymphoma and breast (Sandler el a1.,1985. 1986; Hirayama, private communication). The first three are significant at the 95% level. The combined breast relative'rtsk of 1.4 is significant at only 88%. Higher relative risks for these four sites might be found for direct smoking if epidemiologists used nonpassively Table 7. Summary: U.S. annual deaths from passive smoking. Lung Cancer Other Cancer Hean Disease Total Females: 1. Constant combined relative risk. 992 8599 9768 19359 2. Relative risk declining with age. 911 11165 7602 19678 3. (1) corrected for misclassification. 1232 12280 14995 28507 Males: 1. Constant combined relative risk. 1606 0 17335 18941 2. Relative risk declining with age. 1606 0 18164 19770 3. (1) eorrected for misclassification. 2499 0 22467 24966 Totals for both sexes: 1. Constant combined relative risk. 2598 8599 27103 38.';00 2. Relative risk declining with age. 2517 11165' 25766 39448 3. (1) corrected for misclassification. 3731 12280 37462 53473 Best current estimate, both sexes (rounded). 3000 11000 32000 46000

~ . Table 8. Summary: Deaths per million population in U.S. from passive smoking. . . ' - (based on 239.000.000 U.S. population in 1985). Lung Cancer Other Cancer Heart Disease Total _._,.... Females: 1. Constant combined relative risk. 4.15 35.98 40.87 81.00 2. Relative nsk declining with age. 3.81 46.11 31.81 82.33 3. (I) corrected for misclassification. 5.15 51.38 62.74 119.27 Males: I. Constant combined relative risk. 6.72 0 ' 72.53 79.25 2. Relative risk declining with age. 6.72 0 76.00 82.72 3. (1) corrected for misclassification. 10.46 0 94.00 104.46 Totals for both sexes: 1. Constant combined relative risk. 10.87 35.98 113.40 160.25 2. Relative risk declining with age. 10.53 46.71 107.81 165.05 3. (1) corrected for misclassification. 15.61 51.38 156.74 223.73 Best current estimate, both sexes (rounded). 13 46 134 193 J 256 A Wells exposed never smokers as the referrent category rather than all never smokers as is usually done. Another dif- ference between passive smoking and direct smoking is that the ratio of lung cancer deaths to deaths from other cancer for females or from heart disease for both sexes is much lower in passive smoking than in direct smok- ing. These differences in mortality effects are probably real and reflect differences in chemistry and physics between direct smoking and passive smoking. Environ- mental tobacco smoke is generated in the burning tip of the cigarette at a lower temperature than direct smoke and therefore contains higher proportions of complicated organic compounds that tend to be carcin- ogenic (Brunnemann et al., 1978). More importantly, (see Appendix D) the mainstream smoke, although generated at a particle size of about 0.7 µm, is very concentrated and appears to agglomerate into larger particles. Deposition rates are high, about 80%. De- position occurs primarily in the mouth or in the larger airways of the lung where the particles are cleared rel- atively quickly into the mouth. This material is then swallowed. Some of it may be eliminated and produce no health effects at all or it may cause the digestive type cancers observed. Only a portion of mainstream smoke appears to remain as small particles that can penetrate deeply into the alveolar region. Environ- mental tobacco smoke, on the other hand, is very dilute, with a mass median diameter of about 0.4 µm. Particles in this size range have very low deposition rates, on the order of 10%, but what does deposit does so deep in the alveolar region of the lung where clearance times are longer. Black and Pritchard (1984) estimate that cigarette tar has a 17 hour half-time rate of clearance from the alveolar region, much longer than clearance times from the ciliated parts of the lung, but much shorter than for inert particles. This means that smoke particles are very likely dissolving in the fluids in the alveolar region and are being cleared into the blood and lymph systems for circulation throughout the body. In summary, there are two types of smoking: (a) large particle smoking, or its equivalent, which is the major compon,ent of direct smoking, which results in massive deposition in the mouth and larger airways of the lung, rapid,clearance, cancers of the mouth. central lung and digestive system. and possibly heart disease, and (b) small particle smoking, which is a minor com- ponent of direCt smoking, but the entirety of passive smoking, and which results in low doses deep in the lung, slow clearance, some lung cancer, but primarily other cancers and adverse heart effects. These differfnces in chemistry and physics also ex- plain, at least in part. the rather high mortality observed for passive smoking relative to the deposited dose of particulate. Smoke retention by a passive smoker is only about 1/400 that retained by a direct smoker in a 16 hour day (0.64 mg for the passive smoker per USSG (1986, p. 196) and 240 mg for the direct smoker assum- ing twenty 15 mg tar cigarettes and 80% retention). In comparison, the ratio of lung cancer death rates is about 1/35. For cancers other than lung in females the ratio is about 1/7, for heart disease in females about 1/ 14 and for heart disease in males about 1/3. Preliminary calculations which are shown in Appendix D indicate that the smoke retained deep in the alveolar region may have a dose ratio higher than 1/400, perhaps as high as 1/60. It may be that carcinogenic material that solu- bilizes and clears from the alveoli into the blood may cause not only some of the cancers other than lung that are observed in passive smoking, but also some of the heart disease from passive as well as direct smoking. The hypothesis of Benditt and Benditt (1973) that ar- terial plaques are caused by DNA-modifying agents is receiving increasing support. See, for example, the re- cent work of Pe n tt al. (1986) on cell transforming capability of hu n theroSclerotic plaque DNA and the earlier work ~ A~bert et at (1977) and Penn et al. (1981) on the for atiort of arterial plaques in cockerels with dimethylbe z(a)ahthracene and benzo(a)pyrene. Another possiN~ factor that might help explain the disparate mortaiity tff4cts versus dose is the level of disease susceptab,ility'itti `passive smokers versus direct ~

. 0.dutt moctatity (com passive smoking 257 smokers. The median age for passive smoking death from lung cancer for males is 66 and the deaths con- stitute 0.006% per year of the exposed population. The first 0.006% of ma)e smokers have died of lung cancer by age 46 at which age the lung cancer death rate is doubling every four years. ikt age 66 the smoker lung cancer death rate is doubling about every 13 years. In other words, in passive smoking deaths we are dealing with only the very most susceptible people, whereas in direct smoking most of the victims are much nearer average susceptibility. Similar considerations apply to the other diseases here discussed. A question often raised is that direct smokers are also passive smokers, so why do they not get the passive smoking related cancers. We have already pointed out that the use of nonexposed never smokers as the re- ferrent category for smoker relative risk would increase the apparent risk for smokers. Another possible expla- nation is the probability of competing risks. Most of the highly susceptible direct smokers would have died in their forties or fifties from smoking related disease and would not be available to die of passive smoking related disease in their sixties or seventies. The passive smoking mortality calculated in this study. namely, 46,000, may be low. Repace and Lowrey (1985) calculate lung cancer deaths from passive smok- ing at 4,665, or about 50% higher than our estimate, primarily because of postulated intense exposure at the workplace, a factor not taken into account in this study since the relative risks are based largely on home ex- posure. If Repace and Lowrey are correct, the higher exposure would lead to corresponding increases in deaths from heart disease and other cancer. Also, only ischemic heart disease is considered here. As the all cause data in Appendix B indicate, other cardiovascular diseases and diabetes may be sensitive to environmental tobacco smoke and may increase the total deaths. The new epidemiological studies on passive smoking support the earlier ones and indicate that not only lung cancer, but other cancer and heart disease are serious problems. In fact, lung cancer appears to be only the tip of the iceberg. To be on the safe side public health policy should be to protect nonsmokers from environ- mental tobacco smoke. Acknowledgemenrs - The author is grateful to Dr. T. Hirayama for his data on individual cancer sites and for the details of his "all cause" data. to R. W. Wilson of the U.S. National Center for Health Statistics for data on the smoking status of U.S. residents by S year age inter- vals. to L. Garfinkel for the person years in his 1981 study. to J. M. Samet for data on mate lung cancer in the New Mexico study. to R. C. Brownson for male lung cancer data in the Colorado study, to P. Buffler for her 33+ year data, to Sir John Crofton for abstracts of Lam a+ al. (1987) and Geng tr af. (1987). to P. Reynolds for the number of cases in their study on female cancer, the number of lung canccr cases. and their qualitative results on males, to D. P. Sandier for nonsmoker data on breast cancer. and to S. C. Hunt for enough data from Martin cr at. (1986a) to calculate an all-exposure relative risk. confidence limits and a weighting factor. The author also wishes to thank James Robins. N. A. Dalager. J. M. Samet. W. J. Blot. L. C. Koo. A. H. Wu. G. Pershagen. D. P. Sandier. D. Trichopoulos and J. L. Repace for helpful correspondence and discussion. References Akiba. S.. Kato. H.. and Blot. W. J. (1986) Passive smoking and lung cancer among Japanese women. Cancrr Res 46. 48t1t-4807, Albert. R. D.. Vanderlaan. M.. Burns. F., and Nishizumi. M. (1977) Cancer Res 37, 2232-2235. Balin, B. J. Broadwell. R. D.. Salcman. M.. El-Kalling. M. (1986) Avenues for entry of peripherally administered protein to the central nervou~system in mouse, rat, and squirrel monkey. J. Comp Neurof 1, 260-280. Benditt. E. P. and Bcnditt. J. M. (1973) Evidence for a monoclonal origin of human atherosclerotic plaques. Proc Nail Acad Sci. 70. 1753-1756. Black, A. and Pritchard. J. N. (1984) A comparison of the regional deposition and short term clearance of tar particulate material from cigarette smoke. with that of 2.5 µm polystyrene micro- spheres, Aero><°ot Sci 1A. 224-227. Blot, W. J. and traumeni. J. F. (1986) Passive smoking and lung cancer. J Nar Cancer lnsr 77. 993-1000. Brownson. R. C.. Reif. J. S.. Keefe. T. J.. Ferguson. S. W.. and Pritzl. J. A. (1987) Risk factors for adenocarcinoma of the lung. Am J Epidentlol 125. 25-34. Brunnemann. K. b., Adams. J. D.. Ho. D. P. S.. and Hoffmann. D. (1978) The Influences of tobacco smoke on indoor atmospheres II. Volatile and tobacco specific nitrosamines in main and side- stream smoke and their contribution to indoor pollution, in Pro- eeedings, 4th'Joint Conjirence on Sensing of Environmental Pollutants. N4w Orle,ns, Louisiana. 1977. American Chemical Society. Washington. D.C., 876-880. Buffler. P. A.. Pickle. L. W.. Mason. T. J.. and Contant. C. (19M) The causes of lung cancer in Texas. in Lung cancer: Causes and prevention. W{. Mizell and P. Correa. eds.. pp. 83-99. Verlag Chemie Interrlational, New York. Chan. W. C.. Colbourne. M. 1.. Fung. S. C.. and Ho. H. C. (1979) Bronchial cancer in Hong Kong 1976-1977. B,J Cancer 39. 182- 192. Chan. W. C. and Fung. S. C. (1982) Lung cancer in non- smokers in Hong Kong. in Cancer campaign. Vol. 6. Cancer Ep- idemiotogy, E. Grundmann. ed., pp. 199-202. Gustav Fischer Verlag. Stuttgart, New York. Correa. P.. Pickle. L. W., Fonthan. E.. Lin. Y.. and Haenszel. W. (1983) PassivtL smoking and lung cancer, l.ancer ti. 595-597. Dalager. N. A.. PiCkle. L. W.. Mason. T. J.. Correa. P.. Fontham. E.. Stemhageh, A.. tt at. (1986) The relation of passive smoking to lung cancer. Cancer Res 46. 48(18-i811. Fong, P. (1982)'J'he hazard of cigarette smoke to nonsmokers. J. Biot. Phys 10; 65-73. Freidman. G. O.; Petitti.. D. B., and Bawol. R. D. (1983) Prevalence and correlates of passive smoking. Am J Public Health 73. 401- 405. Fuchs. N. A. (1964) The Mechanics of Aerosols. pp. 46-49. Mac- millan. New York. Garfinkel. L. (1481) Time trends in lung cancer mortality among nonsmokers and a note on passive smoking. J Nat Cancer Inst 66. 1061-1066. GarGnkel. L.. Auerbach. O., and Joubert. L. (1985). Involuntary smoking and lung cancer: A case control study. J Nar Cancer Insr 75.463-469. Garland. C.. Barrett-Connor. E.. Suarez. L.. Criqui. M. H.. and Wingard. D. L. (1985) Effects of passive smoking on ischemic heart disease mortality of nonsmokers. Am J Epidcmiot 121. (.35- 650. Geng. G. Y., l.iang. Z. H.. Zhang. A. Y., and Wu. G. L. (1987. November) On the relationship between cigarette smoking and female lung cancer. Paper presented at the 6th World Conference on Smoking and Health. Tokyo. Gerrity, T. R.. Lee. P. S., Hass. F J.. Marinelli. A.. Werner. P.. and Lourenso, R, V. (1979) Calculated deposition of inhaled particles in the airway generations of normal subjects. J Appl Ph}•siol: Respirar Environ Exercise Phvsiol 47. 867-873. Gillis. C. R.. Hole. D. J.. Hawthorne. V. M.. and Boyle. P. (1984) The effect of environmental tobacco smoke in two urban com- munities in the west of Scotland. EurJ Resp Dis 65. (supplement No. 133). 121-126. Hammond. C. (1966) Smoking in relation to the death rates of one million men and women. in Epidemiological Approaches to the Study of Cancer and Other Chronic Diseases. W. Haenszel. ed.. pp. 127-2p1. U.S. Public Health Service. Bethesda. MD. (Na- tional Cancer Institute Monograph 19). 50986 1067

A. J. Wells . Helsing. K. J. • Sandter. D. P.. Comstock. G. W.. and Chee. E. (1988) Heart discasc mortalitv in nonsmokers living with smokers. Am J 6pidtmiol 127. 915-922. Heyder. J. (1984) Studies of particle deposition and clearance in humans• in Problenu ojlnhatato.,v Toxicuy Studies P. Grosdanoff, tt al., eds.. pp. 155-180. MMV-Medizin-Verlag. Munich. Hitler, F. C.. McCusker. K. T.. Mazumdcr M. M. K.. Wilson. J. D.. and Bone. R. C. (1982) Deposition of sidestream ciagarette smoke in the human respiratory tract. Am Rev Resp Dis 12S• 406-J08. Hirayama. T. (1984a) Cancer mortWity in nonsmoking women with smoking husbands based on a large-scale cohort study in Japan. Prey Med 13. 680-690. Hirayama. T. (1984b) Lung cancer in Japan: Effects of nutrition and passive smoking. in Lung cancer: Causes and prevention. M. Mis- eli. P. Correa, eds.. pp. 175-195. Verlag Chemie International. New York. Hirayama, T. (1987) Passive smoking and cancer: An epidemiological review, in Gann Monograph on Cancer Research 33. Japan Sci- entific Societies Press. Tokyo, pp. 127-135. Humble. C. G., Samet. J. M., and Pathak. D. R. (1987) Marriage to a smoker and lung cancer risk, Am J Pub Health 77, 598-602. lnoue• R. and Hiraysma. T. (1987. November) Passive smoking and lung cancer in women. Paper presented at the 6th World Con- ferena on Smoking and Health, Tokyo. Kabat. G. C. and Wynder. E. L. (1984) Lung cancer in nonsmokers. Cancer 53,1214-1221. Kabat• G. C.. Diech, G. S., and Wynder• E. L. (1986) Bladder cattcer in nonsmokers. Cancer S7, 362-367. Kalandidi. A.. Trichopoulos. D., Hatzakis. A.. Tzannes. S., and Saracci. R. (1987) Passive smoking and chronic obstructive lung disease. Lancet ii. 1325-1326. Knoth. A.. Bohn. H., and Schmidt. F(1983) Passivrauchen als l.un- genkrebs-Ursache bei Nichtraucherinnen. Medizinische Klinik 78. 66-69. Koo. L. C.. Ho. J. H. C.. Saw. D., and Ho. C. (1987) Measurements of passive smoking and estimates of lung cancer risk among non- smoking Chinese females. Inr J Cancer 39. 162-169. Lam. T. H.. Kung,1. T. M.. Wong, C. M.. Lam. W. K.. Kleevens, 1. W. L.. Saw. D.• tt al. (1987) Smoking. passive smoking and histological types in lung cancer in Hong Kong Chinese women. Br ! Cancer 56. 673-678. Lam. W. K. (1985) A clinical and epidemiological study of carcinoma of lung cancer in Hong Kong. M.D. thesis. University.of Hong Kong. Hong Kong. Lee. P. N.. Chamberlain. J., and Alderson. M. R. (1986) Relationship of passive smoking to risk of lung cancer and other smoking- associated diseases. Br J Cancer $4• .97-105. Martin. M. 1.. Hunt. S. C.. and Williams. R. R. (1986a. October) Increased incidence of heart attacks in nonsmoking women mar- ried to smokers. Paper presented at annual meeting of American Public Health Association. Martin. M. J.. Svendsen. K. H., and Kuller. L. H. (1986b. March) Nonsmoking men married to smokers are similar to nonsmoking men married to nonsmokers. Paper presented at the 7th Annual Meeting, Society for Behavioral Medicine. Melandri. C.. Tarroni. G.. Prodi. V.. DeZaiacomo. T.. Formignani• M., and Lombardi. C. C. (1983) Deposition of charged particles in the human airways. J Aerosol Sci 14. 657-669. Miller. G.H. (1984) Cancer. passive smoking and non-employed and employed wives. Wtst J Med 140. 632-635. Mitchell. R. I. (1962) Controlled measurement of smoke particle retention in the respiratory tract. Am Rev Respir Dis 85• 526- 533. National Centei for Hea)th Statistics (1986) Health United States 1986. U.S. Dept. of Health and Human Services, Public Health Service. Hyzittsville. MD, 106-109. National Research Council (1986) Environmental tobacco smoke• measuring exposures and assessing health effects. National Acad- emy Press. Washington. D.C. Office of Technology Assessment (1985) Smoking related deaths and financial costs: Office of Technology Assessment. U.S. Congress. Washington. D.C. Office on Smoking and Health (1979-85) Bib- liography on s,mokitlg and health. U.S. Public Health Service. Rockville, MD. and Smoking and Health Bulletins after 1985. Penn. A.. Batastini. G.. Solomon. J.. Burns. F.. and Albert. R. E. (1981) Cancer $ej 41. 588-592. Penn. A.. Garte. S, J., Warren. L.. Nesta, D., and Mindich. B. (1986) Transforming gene in human atherosclerotic plaque DNA. Proc Nat Acad Sci 83, 7951-7955. Pershagen• 0.. Hn4bec. Z., and Svensson. C. (1987) Passive smoking and lung cancer':in Swedish women. Am J Epidemiol US. 17-.d. Repace. J. L. and kowrey. A. H. (1985) A quantitative estimate of twnsmokeri lung cancer risk from passive smoking. Environ !nr 11. 3-22. Reynolds, P.. Kaplan. 0. A., and Cohen. R. D. (1987, June) Passive smoking and cane,er incidettce: prospective evidence from the Ala- meda County study. Paper presented at the Society for Epide- miologic Research. Amherst. Massachusetts. Rothman. K.1. and "ce. l. D. (1982) Epidemiologic Anal,vsi's with a Programmable Calculator. pp. 5-17. Epidemiology Resources. Chesnut Hill. Massachusetts. Rothman. K. l. (1986) Modern Epidemiology. pp. 139-147. 184-190. Little. Brown, Boston. Russell, M. A. H.. Jarvis. M. 1.. and West. R. 1. (1986) Use of urinary nicotine cohcentrations to estimate exposure and mortality from passive smoking in non-smokers. 8r J Addiction 81. 317- 323. Sandler. D. P.. EversQn, R. B.. and Wilcox. A. J. (1985) Passive smoking in adulthood and cancer risk. Am J Epidemiol 121. 37- 48. Sandler. D. P.. Everson. R. B.. and Wilcox. A. J. (1986) Cigarette smoking and breast cancer. Arrt J Epideamiol 123. 370-371. Stbber, W. (1984) Lung dynamics and uptake of smoke constituents by nonsmokers-a lurvey, Prey Med 13, 589-601. Svettdsen. K. H., Kulk'~ ,[.. H.. and Neaton. J. D. (1987) Effects of passive smoking in the multiple risk factor intervention trial. Am J Epidemiol 126, 70-795. Trichopoulos. D.. Kalandidi, A., and Sparros. L. (1983) Lung cancer and passive smoking: Conclusion of the Greek study. Lancer ii. 677-678. U.S. Surgeon General, (1986) The health consequences of in- voluntary smoking. a report of the Surgeon General. DHHS (CDC) 87-8398, U.S. Public Health Service. Rockville. Mary- land. Vandenbroucke. J. P.. Verheesen, l. H. H., deBruin. A.. Mau- ritz. B. J.. Van der He,ide-Wessel. C., and Van der Heide. R. M. (1984) Active and pas#ive smoking in married coupks: Results of 25 year follow up. Br,btd J 288. 1801-1802. Wald, N. J., Nanchanal. K.. Thompson. S. G., and Cuckle. H. S. (1986) Does breathing" other people's tobacco smoke cause lung cancer? Br Mtd J 293. 1217-1222. Wells. A. J. (1986) Misclassification as a factor in passive smoking risk. Lancet ii, 638. Wu. A. H.. Henderson. B. E.. Pike. M. C.. and Yu. M. C. (1985) Smoking and other risk factors for lung cancer in women. ! Nat Cancer Insr 74. 747-751.