Tobacco Documents Online

-4 Y l - --7,2c~.W-- s~c3 Carrer Causes and C:ontroi. 1997. 8, pp. 57-64 Environmental tobacco smoke and lung cancer mortality in the American Cancer Society's Cancer Prevention Study II Victor M. Cardenas, Michael J. Thun, Harland Austin, Cathy A. Lally, W Scott Clark, Raymond S. Greenberg, and Clark W. Heath, Jr. (Received 8 Apri11996; accepted in revised form 5 August 1996) k LO Q_~_k Environmental tobacco smoke (ETS) has been classified as a human lung carcinogen by the United States Environmental Protection Agency (EPA), based both on the chemical similarity of sidestream and mainstream smoke and on slightly higher lung cancer risk In never-smokers whose spouses smoke compared with those married to nonsmokers. We evaluated the relation between ETS and lung cancer prospectively in the US, among 114,286 female and 19,549 male never-smokers, married to smokers, compared with about 77,000 female and 77,000 male never-smokers whose spouses did not smoke. Multivariate analyses, based on 247 lung cancer deaths, controlled for age, race, diet, and occupation. Dose-response analyses were restricted to 92,222 women whose husbands provided complete information on dgarette smoking and date of marriage. Lung cancer death rates, adjusted for other factors, were 20 percent higher among women whose husbands ever smoked during the current marriage than among those married to never-smokers (relative risk [KR] = 1.2, 95 percent confidence Interval [CIJ = 0.8-1.6). For never-smoking men whose wives smoked, the RR was 1.1 (CI = 0.6-1.8). RLsk among women was similar or higher when the husband continued to smoke (RR = 1.2, CI = 0.8-1.8), or smoked 40 or more cigarettes per day (RR = 1.9, CI =1.0-3.6), but did not increase with years of marriage to a smoker. Most Cis Included the null. Although generally not statistically significant, these results agree with the EPA summary estimate that spousal smoking Increases lung cancer risk by about 20 percent in never-smoking women. Even large prospective studies have limited statistical power to measure precisely the risk from ETS. Cancer Causes and Contml 1997, 8, 57-64 Key words: Lung cancer, environmental tobacco smoke, nonsmokers, United States. Introduction Environmental tobacco smoke (ETS) is a common air and mainstream tobacco smokeZ and on the higher lung pollutant to which many people are exposed. The United cancer risk seen in never-smokers married to smokers in States Environmental Protection Agency (EPA) has clas- 24 of 30 published epidemiologic studies.i From its pooled sified ETS as a known (Group A) human lung carcinogen,l analysis of 11 published US studies, the EPA estimated based on the numerous carcinogens found in both ETS that never-smoking women married to smokers had 19 Drs Cardenas, Austin, and Clark are with the Epidemiology Division, Rollins School of Public Health, Emory University, Atlanta, GA, USA. Dr Thnn, Ms. Lally, and Dr Heath, Jr. are with the Department of Epidemiology and Surveillance Research, American Cancer Society, Atlanta, GA. Dr Greenburg was formerly with the Rollins School of Public Health, Emory University, and is now with the Medical University of South Carolina, Charleston, SC. Address correspondence to Dr Thun, Department of Epidemiology and Surveillance Research, American Cancer Society, 1599 Clifton Road, NE, Atlanta, GA 30329-4251, USA. 0 1997 Rapid Science Publishers Cancer Causes and Control. Vol 8. 1997 57

ti' VM. Cardenas et al percent higher lung-cancer risk than did those married to never-smokers (relative risk [RRJ = 1.19, 90 percent confidence interval = 1.04-1.35). Because of the ubiquity of ETS exposure, this corresponds to approximately 3,000 lung cancers caused annually by ETS among US never- smokers and former smokers, In addition to other adverse effects on respiratory Infections and asthma from ETS in children.' Epidemiologic studies of ETS cannot make precise measurements of RRs in the range of 1.2 for three reasons. First, lifelong exposure to common environmental pollutants such as ETS is extremely difficult to measure precisely, and misclassification of exposure may bias studies ~ towards the null. Second, potential confounding by active - ---~. . - - cigarette smoking is so strong tF~iat thw~~ e~TS association can, _-__ -- --_L -- -- -- ..~~._.-su only be evaluated among lifelong never-mokers. Third, the iow background inc[- ence of lung cancer in lifelong never-smokers severely limits the statistical power of even large epidemiologic studies, particularly prospective studies. The two published prospective studies of lung cancer and ETS in the US included only nine3 and 153' lung cancer deaths among never-smoking women. More than 1,000 expected cases are needed to achieve 80 percent statistical power (1,014 cases assuming RR = 1.2, cc = 0.05, two-sided testing, and 60 percent of never- smoking women exposed). Despite these methodologic difficulties, accumulated evidence from diverse prospective and retrospective epidemiologic investigations is an important component of the empirical basis for public health policy when considered in aggregate and in relation to observations from other disciplines.l The prospective analyses presented here, although limited in statistical power (247 lung cancer deaths among 288,000 married, never- smoking adults), contribute to this total pool of scientific information. Materials and methods Study population Lifelong nonsmokers (never-smokers) in the analyses were drawn from among 676,526 women and 508,576 men enrolled in Cancer Prevention Study II (CPS-II) (Table 1) 5` Subjects were the friends, neighbors, and relatives of more than 77,000 American Cancer Society (ACS) volunteers in al150 states, the District of Columbia, and Puerto Rico. Enrollment was restricted to persons age 30 and older from families where at least one partici- pating member was 45 years old or older. Enrollees completed a four-page baseline questionnaire in 1982 that included personal identifiers, demographic charac- teristics, personal and family history of cancer and other diseases, and various behavioral, environmental, occupa- 58 Cancer Causes and ControL Vol 8. 19H7 tional, and dietary exposures. Subjects' vital status was determined in two ways from the month of enrollment through 31 December 1989. First, the volunteers made personal enquiries in Septem- ber of 1984, 1986, and 1988 to determine whether their enrollees were alive or dead and to record the date and place of all deaths. Second, automated linkage using the National Death Index (NDI) was used to extend follow- up from August of 1988 through 31 December 1989, to conf "um known deaths during the first sixyears of follow-up, and to identify previously unrecognized deaths among the 21,831 (1.8 percent) people lost to follow-up between 1982 and 1988 $ Death certificates were obtained for 96.8 percent of people known to have dled, and cause of death was classified according to the International Classification of Diseases, Ninth Revision (ICD-9) e At completion of mortality follow-up in December 1989, 1,080,689 (91.2 percent) were still living, 101,519 (8.6 percent) had died, and 2,894 (0.2 percetrt) had follow-up truncated on 1 September 1988 because of insufficient data for NDI linkage. Follow-up beyond 1989 was not included in the present analyses because of concern about increasing misclassification of exposures with the passage of time since the baseline interview. Endpoints Deaths from lung cancer were defined as primary carci- nomas of the lung, bronchus, or trachea (ICD-9 code 162), coded as the underlying cause from the death certifcate~ We determined the fraction of these tumors that might be metastatic to the lung rather than primary by comparing death certificates and tumor registry diag- noses for all (29) never-smokers who resided within the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) cancer registries. SEER listed primary lung cancer as the diagnosis for 27 (93.1 percent) of these decedents. For two, the primary site was listed as unknown by SEER. We also reviewed death certificates for 247 never-smokers coded as lung cancer deaths in the spousal cohort. Histologic information was available from 115 certificates. Of these, 80 (70 percent) were classified as adenocarcinomas, the principal cell type seen in other studies of ETS.10 The histologic information was consid- ered too incomplete to support separate analyses by cell type. ETS exposure assessment Information on potential ETS exposure was obtained at baseline only and was not updated during the seven-year follow-up. The CPS-II baseline questionnaire provided two measures of potential exposure. First, active smoking by the current spouse, when combined with age at marriage, reflected potential ETS exposure to cigarettes, pipes, or cigars (from smoking husbands) or cigarettes

ETS and lung cancer only (from wives) during the present marriage. We obtained the information on active smoking directly from the spouse and linked this to the index never-smoker. Spousal smoking was classified by status (never-, ever-, continuing, or former), duration (years of smoking in the current marriage) and intensity (cigarettes per day) at the time of enrollment, and pack-years of potential exposure during the marriage (the product of duration and ciga- rettes per day). Spousal information provided our only measure of the long-term duration of exposure, which in active smokers is the single largest contributor to lung cancer risk."•'Z Fewer than two percent of subjects were excluded from the analysis because of missing data on spousal smoking (Table 1). Self-reported information on ETS exposure was elicited by the question, 'Whether or not you smoke, on the average how many hours per day are you exposed to the smoke of others: at home , at work , in other areas ?' These questions referred to current rather than past exposures and were often left 'blank.' No information was reported In one, two, or three of the self-reported fields for 18 percent, 30 percent, and 26 percent of women and 15 percent, 26 percent, and 13 percent of men, respectively. To determine whether 'blank' values should be considered unexposed or be excluded, we compared the prevalence of ETS exposure at home in CPS-If and in the 1988 National Health Interview Survey (NHIS).13 Prevalence in the two surveys was most similar (difference usually less than five percent within gender and race subgroups) when 'blank' ETS values in CPS-II were considered 'zeros.' Our analyses, therefore, combined the hours of ETS exposure in the three settings, considering 'blank' values to indicate no exposure. Self-reported data on ETS exposure were considered less informative than were the spousal reports because of the many 'blank' values and the lack of information on duration. Exclusions Table 1 shows the number of subjects and deaths from lung cancer that remained In the spousal and self-reported analytic cohorts after exclusions. All analyses excluded ever-smokers, persons whose smoking was unclassifiable, and those who reported extant cancer (except non-mela- noma skin cancer) at enrollment (Table 1). The spousal analyses also excluded persons without a participating spouse, those whose smoking status was unknown, and those whose date of marriage. or whose spouse's age of starting or stopping smoking were missing. Analyses of self-reported ETS exposure excluded index subjects for whom hours of exposure were unquantifiable. Exposed women in the dose-response subanalyses of the spousal cohort (not shown in Table 1) were restricted further to 92,222 never-smoking women married only once to men who smoked only cigarettes and who provided complete information on years of smoking, cigarettes smoked per day, year of quitting, and age at marriage. Analyses We used Cox proportional hazards (PH) modeling14 to compare lung cancer death rates in never-smokers with or without ETS exposure. The principal contrast involved exposure to spousal smoking (never- cfever-, continuing, or former). Women were categorized further according Table 1. Number of Cancer Prevention Study (CPS)-II persons and deaths from lung cancer included in or excluded from the analyses, ACS Women Men Persons Deaths Persons Deaths Total cohort 676,526 2,686 508,576 5,469 Exclusions: Ever smoked 286,869 2,190 365,275 5,173 Smoking unclassified 34,140 95 16,139 137 Lung cancer at baseline 154 38 28 10 Other cancers at baseline 31,172 74 6,012 22 No spouse in CPS-Ila 103,776 113 15,510 18 Unknown exposure in marriagea 17,673 13 3,224 3 Spousal smoking unclassifieda 10,508 13 5,846 9 Self-reported ETS unclassifiableb 42,655 43 10,435 11 Analytic cohorts Spousal 192,234 150 96,542 97 Self-reported 281,536 246 110,687 116 a Excluded from spousal analyses only. b Excluded from self-reported environmental tobacco smoke (ETS) only. Cancer Causes and Control. Vo18. 1997 59

1!M. Cardenas et al to the type of tobacco smoked by their husbands (ciga- rettes only, cigars/pipes only, or mixed), years in marriage to a smoker (1-17, 18-29, 30+), cigarettes per day smoked by the spouse (0, 1-19, 20-39, 40+), and cumulative pack-years of exposure (1-16,17-35, 36+). Dose-response trends could not be examined in men because of insuffi- cient numbers of deaths. We also measured multivarlate RRs according to the total number of hours of self-reported ETS exposure per day (0, 1-2, 3-5, 6+) reported at enrollment. We compared the lung cancer RR in a subgroup of women who had a smoking spouse and also reported ETS at home (n = 33,030) with the rate among women who had neither (n = 41,817), reasoning that this group would have the least misclassification with respect to current exposure. We also tested the proportional hazards assumption during follow-up using univariate survival curves and displaying the Kaplan Meier estimates in a log-log sccale. " Adjustment for confounding Gender-specific PH analyses controlled either for age only (nine categories) or for: age; race (White cf non- White); years of education (< 12, 12+); blue collar employment (most recent or current job); occupational asbestos exposure (yes/no); weekly servings of vegetables or citrus fruit (carrots, squash, corn, green leafy vegeta- bles, cabbage, broccoli, Brussels sprouts, tomatoes, and citrus fruits and juices) (<_ 2,2 -11,12-15,16-20, 21+); total dietary fat intake in quintiles;`s and self-reported history of chronic lung disease (asthma, chronic bronchitis, emphysema, tuberculosis) (yes cf no). A trend test for dose-response treated the categories as equally spaced ordinal variables in a Cox PH model. Statistical power and inference As mentioned, even very large prospective studies have minimal statistical power (or precision) to measure RRs in the range of 1.2 for uncommon illnesses. The statistical power of the spousal CPS-II analyses was 15 percent in women and 10 percent in men (based on the above assumptions, 150 and 97 observed lung cancer deaths in women and men, and 60 percent and 20 percent of women and men exposed). Despite the imprecision of the RR estimates, we assessed whether the risk patterns were consistent with those of active smoking and lung cancer. We hypothesized that: (1) risk caused by ETS should increase as long as the spouse continued to smoke, but the RR should begin to decrease after cessation; (ii) risk in the never-smoker should increase with indices of cumulative spousal smoking; and (iii) risk should be higher for exposure categories that have the least amount of misclassification (i.e., in persons married to smokers who also report current ETS exposure at home). Results Spousal smoking Fifty-nine percent of never-smoking women in CPS-II were married at baseline to a husband who currently (27.6 percent) or formerly (31.8 percent) smoked during the marriage (Tables 2 and 3). Proportionately fewer never- smoking men were married to current (10.3 percent) or former (10.4 percent) smokers. Women potentially exposed to spousal ETS outnumbered men by nearly six to one, due to more men having been excluded because of active smoking (Table 1) and fewer never-smoking men being married to smokers (Table 2). Demographic differences between persons exposed and unexposed to ETS in the spousal analyses were small and reflected gender and socioeconomic correlates of smoking (Table 2). Compared with unexposed women, wives whose husbands continued to smoke were somewhat less educated and more likely to be employed in a blue collar job; women whose husbands had quit smoking were older and less educated but otherwise similar. Compared with unexposed men, those whose wives formerly smoked were slightly younger, more likely to be actively employed in a non-blue collar job, and more likely to consume vege- tables more frequently and fatty foods less often. Women had potentially greater exposure to ETS from spousal smoking than did men. The number of cigarettes per day, years of smoking, and cumulative pack-years was approximately one-third larger in husbands who smoked than in wives who smoked. Table 3 presents age-adjusted death rates and RRs for lung cancer, plus multivariate RRs and 95 percent confi- dence intervals (CI) adjusted forage, race, education, diet, occupational asbestos exposure, and a history of chronic lung disease. Never-smoking women whose husbands ever smoked during the current marriage had 20 percent higher death rates from lung cancer than did those married to never-smokers (multivariate RR = 1.2, CI = 0.8-1.6) (Table 3). The multivariate RR was similar or higher when the husband continued to smoke (RR = 1.2, CI = 0.8-1.8) or smoked cigars or pipes (RR = 1.5, CI = 0.8-2.9). For never-smoking men whose wives smoked, the RR was 1.1 (CI = 0.6-1.8), although this estimate was based on only eight lung cancer deaths. None of the CIs excluded the null (Table 3). Dose-response spousal subcohort Compared with never-smoking women whose husbands did not smoke during the marriage, the lung cancer risk among never-smoking women married to cigarette smokers increased with cigarettes per day and with pack-years of spousal smoking but not with years in marriage to a smoker (Table 4). Women whose husbands smoked 40 or more cigarettes per day had the highest risk (RR = 1.9, 60 Cancer Causes and Control. Vo18. 1997

ETS and lung cancer Table 2. Demographic characteristics of lifelong never-smokers in spousal cohort, ACS Women (n -192,234) ~~ Men (n = 96,542) ~ Ever-smoking Never-smoking spouse spouse Ever-smo spous king Never-smoking e spouse ~~ ~ People (no.) 114,286 77,948 19,549 76,993 People (%) 59.5 40.5 20.2 79.8 Age (mean yrs) 55.1 542 56.1 57.2 White (°Yo) 95.4 952 94.5 95.0 Blue collar worker (%) 8A 7.1 21.3 28.0 Education (% < 12 yrs) 11.1 8.7 9.1 12.2 Dietary vegetables (%)a 1 (least) 3.9 32 4.4 4.2 2 20.3 18.4 24.9 24.9 3 19.6 19.0 212 21.0 4 23.8 24.5 22.6 23.3 5 (most) 24.0 26.8 20.5 20.4 Dietary fat (%) b 1 (least) 13.9 13.3 12.5 10.9 2 20.5 20.0 17.7 15.6 3 20.9 20.8 18.0 17.4 4 19.5 20.4 202 21.0 5 (most) 16.8 17.4 25.2 28.9 Incomplete nutrition data 8.4 8.1 6.4 62 a Vegetable consumption categories defined as <2, 2-11, 12-15, 16-20, 21+ servings per week. b Fat consumption index defined by quintiles estimated from fat content of 18 foods, average portion size for age and gender, and frequency of consumption.12 Table 3. Lung cancer death rates among married lifelong never-smokers in CPS-11 according to smoking by the spouse, ACS Spousal smoking habits~w No. of people No. of fung~ cancer deaths Age-adjusted rate ratloa Multivariate rate ratiosb (Ci)`~ a~ Women Never smoked 77,948 54 1.0 1.0 Ever smoked 114,286 96 1.2 1.2 (0.8-1.6) Continuing smoking Any type 53,139 44 1.3 1.2 (0.8-1.8) Cigarette only 33,371 26 1.3 1.2 (0.8-2.0) Cigar/pipe only 9,794 11 1.5 1.5 (0.8-2.9) Mixed 9,974 7 1.0 1.0 (0.5-2.2) Former smoking Any type 61,147 52 1.1 1.1 (0.8-1.6) Cigarette only 43,563 39 12 1.2 (0.8-1.8) Cigarlpipe only 6,810 7 1.3 1.3 (0.6-2.8) Mixed 10,774 6 0.7 0.7 (0.3-1.7) Men Never smoked 76,993 79 1.0 1.0 Ever smoked 19,549 18 1.0 1.1 (0.6-1.8) Continuing cigarettes 9,492 8 1.0 1.0 (0.5-2.0) Former cigarettes 10,057 10 1.1 1.1 (0.6-2.2) a From Cox PH model adjusted only for age. ~' From Cox PH model adjusted for age, race, education, dietary consumption of vegetables, and total fat, asbestos exposure, blue collar employment, and history of chronic lung disease. C Ci = 95% confidence interval. Cancer Causes and Contro! Vo18. 1997 61

VM. Cardenas et al Table 4. Number of lung cancer deaths, person-years (PY) at risk, and rate ratios (RR) with 95 percent confidence intervals (Cl) among never-smoking women according to various indices of spousal smoking, ACS ~ No. of women -~ Lung cancers PY ~ ~RRb ~~ (CI) Cigarettes per day by spousea 0 46,149 30 333,946 1.0 1-19 11,467 9 83,074 1.1 0.5-2.2 20-39 24,735 22 179,751 1.2 0.7-2.2 40+ 9,871 13 71,618 1.9 1.0-3.6 Years in marriage to smoke 0 ra 46,149 30 344,946 1.0 Trendb P= 0.03 - 1-17 14,794 13 107,681 1.5 0.8-2.9 18-29 15,491 14 112,761 1.5 0.8-2.8 30+ 15,788 17 114,002 1.1 0.6-2.1 Pack-years of exposurea 0 46,149 30 334,946 1.0 Trendb P= 0.5 - 1-16 15,451 10 112,318 1.0 0.5-2.1 17-35 15,569 16 113,119 1.5 0.8-2.7 36+ 15,053 18 109,006 1.5 0.8-2.6 Trendb P 3 0.1 a The referent group includes never-smoking women married to husbands who did not smoke during the marriage. The exposed categories are split into approximate tertiles, and are restricted to never-smokers married to cigarette smokers with complete smoking data, married only once, and with valid information on age at marriage. b From a Cox PH model adjusted for age, race, education, dietary consumption of vegetables and total fat, asbestos exposure, blue collar employment, and history of chronic lung disease. CI = 1.0-3.6, P trend = 0.03). Restricting the dose-response analysis to women whose current husbands continued to smoke in 1982 reduced the number of lung cancer deaths and the trend. With this restriction, women whose husbands smoked 20 to 39 cigarettes per day experienced 10 lung cancer deaths, (RR =1.6, CI = 0.8-3.3); those whose husbands smoked 40+ cigarettes per day had two lung cancer deaths (RR = 0.9, CI = 0.2-4.0, Ptrend = 0.3). Dose-response trends could not be examined in the men because of a statistically insufficient number of lung cancer deaths. Other analyses The association between lung cancer and spousal smoking was not stronger among women whose husbands currently smoked and who also reported ETS exposure at home (RR = 1.2, CI = 0.7-2.2). This was based on only 23 exposed cases and had limited precision. The corre- sponding RRs among men with concordant exposure data could not be examined (two exposed cases). None of the self-reported current ETS exposure measures (any exposure or total hours of exposure) was associated with increased lung cancer risk. Multivariate RRs among women who reported 0, 1-2, 3-5, or 6+ hours of ETS per day in all settings were 1.0, 0.8, 0.7, and 1.1 based on 175, 19, 11, and 31 lung cancers, respectively. Corresponding RRs in men were 1.0, 0.6, 1.0, and 1.3 based on 74, 20, 8. and I41ung cancer deaths, respectively. Graphs of the univariate survival curves associated with spousal ETS exposure indicated that the mortality rate ratios were approximately constant during the seven years of follow-up in CPS II. Further, the assumption of proportional hazards did not appear to be violated in any of the Cox models. Discussion Our principal finding was that never-smoking women married to husbands who ever smoked during the marriage had 20 percent higher death rates from lung cancer than those married to never-smokers. Risk was similar or increased when the husbands continued to smoke, smoked more cigarettes per day, or exceeded 35 pack-years of cigarettes during the marriage. Although only one dose-response trend was statistically significant, the magnitude and direction of risk were similar to the pooled value from other US studies,' and paralleled tem- poral patterns seen with active smoking, risk increasing with sustained cumulative (pack-years) of exposure and the RR decreasing with cessation. These results should be interpreted in the context of three lines of evidence cited by the EPA and other scien- tific groups as support for the carcinogenicity of ETS.'"z•`r"" First, most of the 3,800 chemicals found in mainstream tobacco smoke (MS) also occur in sidestream smoke (SS). ETS contains both SS and exhaled MS. At 62 Cancer Causes and Control. Vo18. 1997

ETS and lung cancer least 42 chemicals identified in MS and SS are known to cause cancer in experimental animals and/or humans.2 MS accounts for 87 percent of lung cancer in the general population. No threshold is seen in the dose-response relation between active smoking and lung cancer; extrapo- lation of risk from active smokers to ETS predicts risk estimates in the rage of an RR between 1.03 and 1.36.`$ Second, biologic markers of Inhaled tobacco smoke show that never-smokers exposed to ETS absorb and metabolize measurable amounts of tobacco smoke.''z''13•" Average concentrations of nicotine and its metabolite cotinine in ETS-exposed urban never-smokers are about 0.1-1.0 percent of those seen in active cigarette smokers, although certain components of tobacco smoke are produced disproportionately in SS.' A tobacco-specific lung carcinogen 4-(methylnitrosamino)-1-(3-pyridy)-1- butanone (NNK) also has been measured in the urine of ETS-exposed men19 and occurs at higher concentrations in sidestream than in mainstream smoke.'s NNK causes pulmonary adenocarcinoma in rodents and may contrib- ute to the excess of lung adenocarcinomas in ETS-exposed never-smokers.19"?2 Third, most published epidemiologic studies of ETS find higher lung cancer risk among never-smoking women whose husbands smoke than among those married to never-smokers. The EPA report discussed 30 published studies encompassing more than 3,000 lung cancer cases from eight different countries.i Twenty-four (80.0 percent) of these studies found higher risk of lung cancer among wives whose husbands ever smoked; nine were statistically significant. Since the EPA report there have been four additional or expanded US studies,1D,zs"2s plus three new studies in China.2°"Z8 The new US studies, counting CPS-II, add over 1,2001ung cancer cases studied in never-smokers. Their results are generally consistent despite low statistical power and large variations in design, location, and analytic methods in individual studies. The CPS-II exposure data were much less detailed than in case-control studies designed specifically to examine the ETS hypothesis.t0'u'x4 Although information on spousal smoking was our best and only indication of long-term potential ETS exposure, this, too, was imper- fect. We lacked information on spousal smoking during previous marriages, smoking by other family members or housemates, proximity to the smoker, changes in the number of cigarettes per day smoked by the spouse, and childhood exposure to ETS. No biologic measurements were obtained to validate absorption of ETS. Our analyses could not integrate potential exposure to a smoking spouse with past or present exposure in other settings. The resultant random misclassification of ETS exposure could bias our results towards the null. Even large prospective studies such as CPS-II have limited statistical precision (wide confidence intervals) when measuring uncommon diseases in a small fraction of the population. Because only two percent of men and eight percent of women in CPS-II were never-smokers married to currently smoking spouses, we traced 676,526 women for seven years to f'md 1501ung cancer deaths in the subgroup of Interest. By contrast, the largest retro- spective study10 identified 653 incident lung cancers in married, never-smoking women in five US cities with a base population of approximately 18,500,000 women. Most other studies of ETS have fewer cases of lung cancer in never-smoking women than CPS-IL The conclusion of some small studies does not always appropriately reflect their imprecision. For example, a recent hospital- based case-control study of 41 lung cancer cases in never-smoking men (RR = 1.60, CI = 0.67-3.82) and 117 cases in women (RR = 1.08, CI = 0.60-1.94)25 concluded 'little indication of an association' between ETS and lung cancer, despite the overlap in confidence Intervals between the study results and the EPA pooled estimate.i CPS-II avoids several biases that have been hypothe- sized to weaken retrospective studies." Spouses described their smoking habits at baseline, before lung cancer had been diagnosed in their partners. This avoids recall bias (awareness of disease influencing the reporting of ETS) and information bias (deceased cases being described by next-of-kin but controls describing them- selves). Since the analysis identified only married couples, no bias was introduced as a result of married and unmar- ried persons describing ETS exposure differently.30 Small demographic differences were apparent between persons exposed or unexposed to spousal smoking in CPS-II. Adjusting for age, race, vegetable and total fat consump- tion, education, occupational exposure to asbestos, and prior lung disease had minimal effect on the association with ETS. Our results pertain largely to lung cancer among never- smoking women married to smoking men. Too few never-smoking men were married to smokers and insuf- ficient information is available on their ETS exposure outside the home to interpret the data on CPS-II men. Because CPS-II was limited to adults, it also lacks infor- mation on adverse ETS-related effects in children such as pneumonia, bronchitis, asthma, and middle ear effusions. Such nonmalignant effects may greatly outnumber lung cancer deaths. EPA estimated that ETS caused approxi- mately 3,000 lung cancer deaths annually among US never-smokers (assuming RR = 1.19) and former smokers, compared with 150,000 to 300,000 cases of bronchitis and pneumonia in infants and children up to 18 months of age, and 200,000 to 1,000,000 children whose asthma would be worsened by ETS.' In summary, our findings contribute to the overall evidence that ETS exposure from smoking spouses may adversely affect lung cancer risk in never-smoking women. Cancer Causes and Control. Vol 8. 1997 63

VM. Cardenas et al Acknowledgements - The authors are grateful for the cooperation of the principal investigators of NCI SEER Cancer Registries, and Dr Christina Park, from CDC's National Center for Health Statistics for prompt re- p1Ies to our inquiries. The authors also thank Audrey Earles for preparation of the manuscript, Mohan Nam- boodiri, Heidi Miracle-McMahill, Diane Terrell, and Dena Myers for programming support, and Dollie D. Daniels and Luis Escobedo for their dedicated editing. Dr Cardenas is recipient of WK Kellogg Foundation and Emory University financial awards. References 1. US Dept. of Health and Human Services, EnvIronmentai Protection Agency. Respiratory Health Effects of Passive Smoking: Lung Cancer and Other DLsoniers. Smoking and Tobacco Control Monograph 4. Washington DC: DHHS, 1993; NIH Pub. No. 93-3605. 2. International Agency for Research on Cancer. Tobacco Smoking: Lyon, France: IARC, 1986; IARC Monogr Eval Cancinog Risk Chem Humans, Vol. 38. 3. Butler WJ. Comments submitted to Environmental Protection Agency on its draft report, Health Effects of Passive Smoking. Assessment ofLung Cancer in Adults and Respiratory Disorders in Children. Washington DC: EPA/60016-90/006A, May 1990. 4. Garfinkel L. T'une trends in lung cancer mortality among nonsmokers and a note on passive smoking. JNCI 1985; 75: 463-9. 5. Steiiman SD, Garfinkel L. Smoking habits and tar levels in a new American Cancer Society prospective study of 1.2 million men and women. JNCI 1986; 76: 1057-63. 6. Peto R, Lopez AD, Boreham J, Thun M, Heath CW. Mortality from Smoking in Developed Countries 1950- 2000: Indirect Estimation from National Vital Statistics. Oxford, UK: Oxford University Press, 1994. 7. Thun MJ, Day-Lally C, Myers DG, et al. Trends in tobacco smoking and mortality from cigarette use in Cancer Prevention Studies 1 (1959-1965) and II (1982-1988). In: National Cancer Institute, Smoking and Tobacco Control, Monograph 6: Changes in Cigarette-Related Disease Risks and Their Implication for Prevention and Control. Wash- ington, DC: National Institutes of Health, in press. 8. Calle EE, Terrell DD. Utility of the National Death Index for ascertainment of mortality among Cancer Prevention Study II participants. Am JEpidemiol 1993; 137(2):235-41. 9. World Health Organization. International Classification of Diseases. Ninth Revision. Geneva, Switzerland: WHO, 1979. 10. Fontham ETH, Correa P, Reynolds P, etal. Environmental tobacco smoke and lung cancer In nonsmoking women: A inulticenter study. JAMA 1994; 271: 1752-9. 11. Doll R, Peto R. Cigarette smoking and bronchial carci- noma: Dose and time relationships among regular smokers and lifelong non-smokers. J Epidemiol Commun Health 1978; 32: 303-13. 12. Peto R. Influence of dose and duration of smoking on lung cancer rates. In: Zaridze D, Peto R, eds. Tobacco: A Major International Health Hazard. Lyon, France: International Agency for Research on Cancer,1986; IARC Sci. Pub. No. 74: 23-33. 13. National Center for Health Statistics. National Health Interview Survey, 1988, Occupational Health Supplement Data. (Wonder Tape DSN = CC36.NHIS88.OCHealth). 14. Cox DR. Regression models and life tables (with discus- sion). J Royal Stat Soc, Series B 1972; 34: 187-220. 15. Thun MJ, Calle EE, Namboodiri MM, et al. Risk factors for fatal colon cancer in a large prospective study. JNCI 1992; 84: 1491-500. 16. National Research Council. Environmental Tobacco Smoke: Measuring Exposures and Assessing Health Effects. Washington, DC: National Academy Press, 1986. 17. US Department of Health and Human Services. The Health Consequences of Involuntary Smoking: A Report of the Surgeon General. Washington, DC: US Government Print- ing Oflice, 1986. 18. US Department of Health and Human Services. Reducing the Health Consequences of Smoking: 25 Years of Progre.cs. A Report of the Surgeon GeneraL Washington, DC: US Govertunent Printing Office, 1989. 19. Hecht SS, Carmella SG, Murphy SE, Akerkar S, Brunne- mann KD, Hoffman D. A tobacco-specific lung carcinogen in the urine of men exposed to cigarette smoke. N Engi J Med 1993; 329: 1543-6. 20. Hoffmann D, Hecht SS. Nicotine-derived N-nitrosamines and tobacco-related cancer: current status and future directions. Cancer Res 1995; 45: 935-44. 21. Hecht SS, Hoffmann D. Tobacco-specific nitrosamines, an important group of carcinogens in tobacco and tobacco smoke. Carcinogenesis 1988; 9: 875-84. 22. Rivenson A, Hoffmann D, Prokopczyk B, Amin S, Hecht SS. Induction of lung and exocrine pancreas tumors in F344 rats by tobacco-specific and Areca-derived N-nitrosami- nes. Cancer Res 1988; 48: 6912-7. 23. Brownson RC, Alavanja MCR, Hock ET, Loy TS. Passive smoking and lung cancer in nonsmoking women. Am J Public Health 1992; 82: 1525-30. 24. Stockwell HG, Goldman AL, Lyman GH, et al. Environ- mental tobacco smoke and lung cancer risk In nonsmoking women. JNCI 1992; 84: 1417-22. 25. Kabat, GC, Steilman SD, Wynder EL. Relation between exposure to environmental tobacco smoke and lung cancer in lifetime nonsmokers. Am J Epidemiol 1995; 142: 141-8. 26. Du YX, Cha Q, Chen YZ, Wu JM. Exposure to environ- mental tobacco smoke and female lung cancer in Guangzhou, China. Proceedings of Indoor Air '93, Vol. 1. International Conference on Indoor Air Quality and Cli- mate. Helsinki, Finland, July 4-8, 1993: 511-6. 27. Lui Q, Sasco AJ, Riboli E, Hu MX. Indoor air pollution and lung cancer in Guangzhou, People's Republic of China. Am J Epldemiol 1993; 137: 145-54. 28. Wang FL, Love EJ, Liu N, Dal XD. Childhood and ado- lescent passive smoking and the risk of female lung cancer. Int J Epidemlol 1994; 23: 223-30. 29. Lee PN. Environmental Tobacco Smoke and Mortality. Basel, Switzerland: S. Karger, 1992. 30. Lee PN. Re: Environmental tobacco smoke and lung cancer in nonsmoking women [Letterj. JNCI 1993; 85: 748. 31. Layard MW. Re: Environmental tobacco smoke and lung cancer in nonsmoking women [Letter] . JNCI1993; 85: 748-9. 32. Gross AJ. The risk of lung cancer In nonsmokers in the United States and its reported association with environ- mental tobacco smoke. J Clin Epidemiol 1995; 48: 587-98. 64 Cancer Causes and Control. Vot 8. 1997