Tobacco Documents Online

V._ .....,.,.~.....e. ._.......:~::~i~~.. . Environmental Tobacco Smoke and Lung Cancer in Nonsmokers: Does Time since Exposure Play aRo1e? Fre a37~ yberg,.,Yeronica Agrenius,2 Katharina Svartengren,3 Christer Svensson,' arld Goran Pershagenl We conducted a population-based case-control study in Stock- holm during 1989-1995 to investigate the risk of lung cancer from exposure to environmental tobacco smoke. The study base consisted of persons above 30 years of age resident in Stockholm County who had never smoked regularly (that is, one cigarette or more daily for 1 year). Cases of lung cancer were identified at the three major county hospitals responsible for diagnosis and treatment of lung cancer. A total of 124 cases (35 men and 89 women) and 235 population controls (72 men and 163 women) participated. The never-smoking status was validated by interviews with next-of-kin. The relative risk associated with ever-cohabiting with a smoking spouse was 1.2 [95% confidence interval (CI) = 0.7-1.91. Ever-exposure at work to environmental tobacco smoke carried a relative risk of 1.6 (95% CI = 0.9-2.9). Risks tended to be more elevated in high-exposure groups and with recent exposures. Both sources of environmental tobacco smoke seemed important, and con- siderable misclassification of total exposure occurred for each variable used separately, in particular for the less common spousal exposure. For those currently exposed to environmen- tal tobacco smoke from the spouse, at work, or both, the relative risk was 2.6 (95% CI = 1.0-6.5). Our data imply that information from major sources of environmental tobacco smoke should be combined to avoid important misclassifica- tion and that timing of exposure should also be taken into consideration. (Epidemiology 1998;9:301-308) Keywords: lung cancer, environmental tobacco smoke, passive smoking, spouse, work, time since exposure, misclassification of exposure, case-control study. Exposure to environmental tobacco smoke (ETS) is common and recognized as a possible risk factor for lung cancer in nonsmokers. A 1992 review by the U.S. En- vironmental Protection Agency summarizes 31 studies of ETS and lung cancer'; since then, several new studies have become available.2-9 Few studies include male non- smokers. Fairly consistent results are observed in relation to ever living with or being married to a smoker, the most common exposure measure. Many studies show dose-response relations for quantitative variables, such as amount smoked by the spouse.lo,' 1 Results for duration of exposure are less consistent, although risks are mostly raised in the category with longest durations.lo,i1 Other dose metrics include cumulative pack-years of expo- sure3,4,6.iZ•'3 and smoke-years',',5 (sum of years from dif- ferent sources). It is unclear which metric best represents relevant biological exposure in various settings. Further- From the Divisions of ' Environmental Epidemiology, Institute of Environmental Medicine, Karolinska Institute, and 'Respiratory Medicine, Thoracic Clinics, Karolinska Hospital, Stockholm; 3Division of Respiratory and Allergic Diseases and 'Oncological Department of Southern Stockholm, Huddinge University Hospitat, Huddinge, Sweden. Addresa reprint requests to: Fredrik Nyberg, Division of Environmental Epide- miology, Institute of Environmental Medicine, Karolinska Institute, Box 210, SE-171 77 Stockholm, Sweden. Submitted June 9, 1997; final version accepted December 10, 1997. 0 1998 tiy Epidemiology Resources Inc. more, variables representing ETS exposure from only one source, such as the spouse, may reflect total exposure inadequately, resulting in exposure misclassification. Cessation of ETS exposure has been studied little, despite the well documented decrease in excess relative risk of lung cancer among smokers after quitting smok- ing.'4'!7 The use of variables that do not address timing of exposure may further increase misclassification of ETS exposure, particularly in populations where exposure has ceased for many persons because of smoking cessation in their environment. Differences in urbanization, occasional smoking, and life-style factors, in particular diet, between subjects exposed and unexposed to ETS have been proposed as explanations for the observed associations between ETS and lung cancer,is,19 underlining the importance of con- sidering these potential confounding factors.' Inadver- tent inclusion of cases and controls who are misclassified smokers, and may introduce confounding, is another concern.t•20 Misclassification studies indicate that this problem is unlikely to cause major bias,21 but evidence from within studies of ETS and lung cancer is limited.I,zz We investigated the relation between ETS exposure and lung cancer among both male and female never- smokers in a Swedish population. We gave special at- tention to validation of never-smoking status, a detailed exposure history regarding ETS from different sources, and evaluation of important confounders. In part, the 301

~ 302 NYBERG ET AL Epidemiology May 1998, Volume 9 Number 3 I study contributes to a collaboration coordinated by 'the International Agency for Research on Cancer. Subjects and Methods STUDY SUBJECTS The study base was composed of Swedish-speaking per- sons 30 years or older who were resident in Stockholm County from October 1, 1989, to September 30, 1995, and who were in a physical and mental condition per- mitting a 1-hour interview. We excluded those who had been regular smokers, defined as having smoked more than 1 cigarette per day, 10 cigarettes per week, or 40 cigarettes per month; or 4 cigarillos per week, 3 cigars per week, or 4 pipes per week for 1 year or longer. We conducted the study in Stockholm County (about 1.7 million inhabitants in 1995). Three hospitals, Karo- linska Hospital, Huddinge Hospital, and Sodersjukhuset, are mainly responsible for diagnosing lung cancer cases in this population. As a rule, a person suspected of having or diagnosed with lung cancer is referred to a lung department at one of these hospitals for further medical investigation, including bronchoscopy and treatment. For case enrollment, we required a microscopically confirmed diagnosis or unambiguous chest radiographs and a typical clinical course. The lung departments were regularly monitored for eligible cases. We screened all persons presenting with suspicion of lung cancer for smoking status (present, former, or never regular smok- er), and all never-smokers were followed until lung can- cer was diagnosed or excluded. For 96% of cases, we retrieved histologic or cytologic slides, which one pulmonary pathologist reviewed to validate the original diagnosis according to the World Health Organization histologic typing of lung tumors.13 Controls from the Stockholm County population reg- ister were frequency-matched approximately 2:1 to cases, in strata defined by gender, age (30-49 years, 50-69 years, 70 years and older), and three hospital catchment areas. Accrual of controls started in March 1992, matched to the initial 33 cases. New controls were selected every 6 months based on the current case dis- tribution. For controls sampled in 1995, matching was only for gender and age, because administrative changes in the health care system had blurred the catchment areas. We approached never-smoking patients when a diag- nosis of lung cancer had been established. We contacted controls by mail, or by telephone if they did not respond, and asked them to participate if they fulfilled the never- smoking screening criteria. If possible, a personal inter- view was scheduled; otherwise, a telephone interview was conducted. We selected all cases (N = 124) and every second control (N = 118) for validation of the never-smoking status of the subject and ETS exposure from the spouse, using a short next-of-kin interview covering the smok- ing habits of the subject and his or her spouse. The subject was not present at this interview. EXPOSURE INFORMATION A trained physician or nurse interviewed subjects using a structured questionnaire. All six interviewers ques- tioned both cases and controls. Two main interviewers performed 76% of the interviews. The questionnaire contained questions regarding oc- casional smoking, a residential history including ad- dresses and building characteristics, a lifetime occupa- tional history, and a food frequency assessment of foodstuffs rich in retinol, carotene, and vitamin C. We assessed exposure to ETS using a core questionnaire developed on the basis of a study on urinary cotinine and ETS exposure?4 It covered childhood exposure, domes- tic exposure from spouse and other cohabitants, and exposure at all work places, at other places, and in vehicles. Exposure to known or suspected occupational lung carcinogensZS was evaluated based on all occupa- tions in the working history, classified as to occupational code (ISCO-68)26 and industrial code (IS1C-71)." STATISTICAL METHODS We constructed ETS exposure variables that consider different sources and environments, as well as time of exposure. We also performed standard analyses using dichotomous or categorical variables for single-source ETS exposure. Numbers of pipes and cigars smoked were converted to cigarette-equivalents using a factor of three. We estimated relative risks (RR) and 95% confidence intervals (Cl) by odds ratios from multiple uncondi- tional logistic regression, adjusted for matching vari- ables: gender, age in five categories (30-49, 50 59, 60-69, 70-79, and ?80 years), and catchment area (three categories). We adjusted all results presented for potential confounders: occasional smoking (three cate- gories), vegetable consumption (three categories), de- gree of urban residence during the past 35 years (three categories), and years of work in risk occupations (con- tinuous). Categorical variables were based on either natural cutpoints (for example, age), or cutpoints from the distribution among controls (25th, 50th, 75th, and 90th percentiles). Results SUBJECT CHARACTERISTICS We identified 145 eligible lung cancer cases among never-smokers. Of these, 8.3% were identified at autopsy or died before interview, and 6.2% refused participation (total nonresponse rate = 14.5%), resulting in 124 en- rolled cases. We were unable to contact or locate 6.1% of potential controls from population registers. These persons may have been smokers or never-smokers, eligible or not eligible for the study. The subsequent screening of con- tacted persons gave a nonresponse rate of 17.1% among never-smokers, resulting in 235 never-smoking controls. Table 1 shows some characteristics of the subjects. The women were slightly older on average than the men (67.1 and 64.7 years for cases and 67.3 and 64.4 years for

Epidemiology May 1998, Volume 9 Number 3 ETS AND LUNG CANCER 303 TABLE 1. Distribution of Selected Characteristics in Lung Cancer Cases and Control Subjects Cases (N = 124) Controls (N = 235) Variable Number %* Number % Gender Male 35 28.2 72 30.6 Female 89 71.8 163 69.4 Age (years) 30-49 12 9.7 26 11.1 50-59 20 16.1 53 22.5 60-69 34 27.4 48 20.4 70-79 42 33.9 61 26.0 ?80 16 12.9 47 20.0 Marital status Ever-married 111 89.5 221 94.0 Never-married 13 10.5 14 6.0 Duration of occupational history Never employed 4 3.2 4 1.7 <10 years 5 4.0 16 6.8 2:10 years 115 92.7 215 91.5 Study interview Personal 111 89.5 163 69.4 By telephone 13 10.5 72 30.6 Hospital/catchment area Huddinge Hospital 48 38.7 112 47.7 Karolinska Hospital 54 43.6 76 32.3 Sbdersjukhuset Hospital 22 17.7 47 20.0 Validation interviewj' Subject refused 1 0.8 5 4.2 Next-of-kin refused 2 1.6 2 1.7 Other nonresp onse$ 6 4.8 3 2.5 Interview with spouse 63 50.8 64 54.2 Interview with other 52 41.9 44 373 relative * Percentage of subjects with data available. t Only 118 of the controls were eligible for the validation interview (see text). $ No relative alive, only relative lives abroad, relative too ill, etc. controls, respectively). Few subjects had never been married. All but one female lung cancer case were histologi- cally (79.0%) or cytologically (20.2%) confirmed. Ade- nocarcinomas were the most common (66.7%); other tumor types included squamous cell carcinomas (9.8%), carcinoids (13.8%), large (3.3%) and small cell cancers (1.6%), and unspecified "malignant cells" or "non-small cell cancer" (4.9%). The pathology review agreed for 77.5% of reviewed cases. Of the remaining tumors, some received more specific classifications; for example, un- specified or large cell cancer was reclassified as adeno- carcinoma (7.5%), others were reclassified from squa- mous cell to adenocarcinoma or vice versa (9.2%) or otherwise reclassified (3.3%); and 2.5% could not be classified from review slides. The next-of-kin validation regarding ever-smoking status of the subject was concordant in 99.1% for cases and 97.2% for controls. For the case and three controls who reported occasional smoking whereas next-of-kin reported "daily or almost daily" smoking, there was no major contradiction regarding total amount smoked, and all smoking was at least 18 years in the past. We con- sidered these discrepancies inconsequential and retained all subjects in the analyses. The validation of exposure reporting showed high agreement on spousal smoking status (>93%, kappa >0.87) and on cumulative amount smoked by the spouse (Pearson's rank correlation 0.75 for reported smokers, 0.92 for all subjects).2' ENVIRONMENTAL TOBACCO SMOKE EXPOSURE The traditional exposure variable "ever living with a smoking spouse" gave a relative risk for lung cancer of 1.2 (95% CI = 0.7-1.9) (Table 2). For men the RR was 2.0 (95% CI = 0.7-5.4), whereas for women it was only 1.1 (95% CI = 0.6-1.9). Restriction to subjects below 70 years of age increased the relative risk to 1.4 (95% CI = 0.7-2.7) for both genders combined. The estimates for reporting actual exposure to a smoking spouse were somewhat lower. For work exposures, the results were similar in both genders, with a combined RR for ever- exposure of 1.6 (95% Cl = 0.9-2.9). No clear increase in risk was associated with dichotomous variables for other indoor exposures or exposure in vehicles, both of which were minor sources of ETS exposure, nor for childhood exposure to father's or mother's smoking, for which some subjects had missing data.

304 NYBERG ET AL Epidemiology May 1998, Volume 9 Number 3 TABLE 2. Relative Risks for Lung Cancer Associated with Basic Dichotomous Variables Representing Environmental Tobacco Smoke Exposure from One Exposure Source at a Time Men Women Both Genders Men Women Variable Cases Controls Cases Controls RR* 95% CI* RR* 95% CI* RR* 95% CI* Spouse smoker Nevert 22 53 39 71 1 1 1 Ever 13 19 50 92 1.17 0.73-1.88 1.96 0.72-5.36 1.05 0.60-1.86 Reported exposure to smoking s pouse Nevert 23 53 43 74 1 1 1 Ever 12 19 46 89 1.05 0.65-1.68 1.64 0.59-4.51 0.94 0.53-1.67 Exposed at work Nevert 5 14 22 55 1 1 1 Ever 30 58 67 108 1.61 0.91-2.85 1.89 0.53-6.67 1.57 0.80-3.06 Exposed in other indoor locatio ns Nevert 24 50 73 125 1 1 1 Evet 11 22 16 38 0.94 0.54-1.63 1.31 0.50-3.38 0.90 0.44-1.86 Nosed in vehicles (non-work-related) everfi 29 63 86 153 1 1 1 Ever 6 9 3 10 0.98 0.41-2.3 7 1.71 0.49-5.98 0.41 0.09-1.75 Childhood exposure to smoking father$ Nevert 9 31 46 75 1 1 1 Ever 20 34 39 73 1.02 0.63-1.66 1.90 0.69-5.23 0.76 0.42-1.37 Childhood exposure to smoking mothert Nevert 9 31 46 75 1 1 1 Ever 6 8 4 13 0.72 0.28-1.87 0.90 0.14-6.00 0.29 0.07-1.14 * Relative risk (RR) and 95% confidence interval (Cl) estimated by odds ratios obtained in multiple logistic regression. Adjusted for matching variables gender, age, and catchment area, as well as occasional smoking, vegetable consumption, degree of urban residence, and years of exposure to risk occupations. t Refetent category. $ Total number of cases or controls is lower owing to missing data for some subjects for the respective exposure variables. We analyzed low and high exposure to spousal ETS using several metrics (Table 3). The risks were elevated in the high-dose group for all variables among men and for both genders combined, and mostly for women. Du- ration alone was not a good predictor of risk, particularly among women. High exposure by cumulative measures (1 "pack-year" is 365 packs, equivalent to I pack per day for 1 year; 1"hour-year" is 365 hours, or 1 hour per day for 1 year) was associated with increased RRs in both genders. Exposure to spousal ETS within the last 15 years showed increased RR among men (2.3; 95% CI = 0.7-7.5), but not women, and the RR for both genders combined was only 1.1 (95% CI = 0.6-1.9). In contrast, the RR for such recent exposure to ETS at work was 2.1 (95% CI = 1.1-4.0) and was consistent among men and women. We obtained similar results for duration or weighted duration of ETS exposure at work (Table 3). Against this background, we examined the exposure distributions in the study base. Among all controls, ever-exposure to ETS from the spouse occurred in 26% of men and 55% of women. The proportion of these exposed controls who were exposed recently, that is, in the last 10 years, however, was 42% among men but only 30% among women. Ever-exposure to ETS at work was more common (men 8 1 % and women 66%), and a larger proportion of the exposed were exposed recently (men 50%, women 43%). Given the problems inherent in treating the two major sources of ETS exposure as independent risk in- dicators, we combined them. Time since last exposure to either major source was associated with elevated relative risks for recent time periods (Table 4), reaching 2.6 (95% CI = 1.0-6.5) for subjects currently exposed. For years of exposure to one or both major sources, the trend in risk was not apparent in the highest category. Expo- sure periods from the two sources may, however, overlap to a varying extent. Another approach is to consider the longer of the two durations more indicative of the total exposure. This analysis showed an elevated risk for sub- jects in the top decile of exposure from one or both sources. Cumulative time-weighted duration avoids the conceptual problem of overlap, because the hours of exposure are not concurrent. Despite some random fluc- tuations, the dose-response gradient is clearer, implying that this may be a useful single variable for representing ETS exposure (Table 4). Nevertheless, with two vari- ables for different facets of ETS exposure, time since end of exposure appears to be a more consistent indicator of risk in our study base than variables based on duration (Table 5). We present all results adjusted for occasional smoking, occupational exposure, urban residence, and diet. We further evaluated confounding by these factors in various models, using different categorical and continuous met- rics, without finding any indication of confounding. Discussion The biologically relevant exposure to ETS with regard to lung cancer is probably determined by duration, inten- sity, timing, and other factors. Previous studies have used "smoking spouse" as an exposure measure and often have included some measure of intensity or cumulative expo- sure. Exposure time window has seldom been addressed, despite evidence from active smoking that indicates that recent exposure is important.ls-" Such analytic strate- gies presuppose continued exposure with moderate in- tensity variation, or at least exposure into the relevant

Epidemiology May 1998, Volume 9 Number 3 ETS AND LUNG CANCER 305 TABLE 3. Relative Risk of Lung Cancer Associated with Reported Exposure to Environmental Tobacco Smoke from Spouse or Work in Low- and High-Exposure Groups, Men and Women Combined Variable Cases Controls RR* 95% Cl* Average daily reported exposure from spouse Unexposedt 66 127 1 <10 cig/day#,§ 40 83 0.96 0.57-1.61 >10 cig/day 15 24 1.16 0.55-2.45 Total duration of exposure to spousal ETS Unexposedt 66 127 1 <30 years§ 39 74 1.01 0.60-1.70 '30 years 19 34 1.14 0.56-2.29 Total weighted duration of exposure to spousal ETS ("hour-years")1f Unexposedt 66 127 1 <90 HY§ 36 84 0.85 0.50-1.44 'i 90 HY 16 23 1.25 0.59-2.66 Cumulative reported exposure to spousal fiTS (pack-years smoked in subject's presence)JI Unexposedt 66 127 1 <9 PY#§ 35 82 0.84 0.49-1.43 ;-_9 PY 20 25 1.53 0.76-3.09 Total duration of exposure to ETS from work Unexposedt 27 69 1 <30 years§ 66 130 1.40 0.76-2.56 Z-_30 years 31 36 2.21 1.08-4.52 Total weighted duration of exposure to ETS from work ("hour-years")[J Unexposedt 27 69 1 <30 HY§ 57 120 1.27 0.69-234 >30 HY 40 45 2.51 118-4.93 Total number of cases or controls varies slightly between analyses owing to missing data for some subjects for the respective exposure variables. * Relative risk (RR) and 95% confidence interval (CI) estimated by odds ratios obtained in multiple logistic regression. Adjusted for matching variables gender, age, and catchment area, as well as occasional smoking, vegetable consumption, degree of urban residence, and years of exposure to risk occupations. t Referent category. t Cigarette-equivalents for pipe or cigar/cigarillo smoke exposure. $ Cutoffs are rounded values to separate into the high-exposure group approximately the top 25% of exposed controls. U One "pack-year" is 365 packs, or the equivalent of 1 pack per day for 1 year; 1"hour-year" is 365 hours, or the equivalent of 1 hour per day for 1 year. time window, which may have been reasonable assump- tions for ETS in many settings. The smoking prevalence in many western societies, however, has declined among men over the last decades but has increased among women.29 As reflected in our data, women ever married to a smoking spouse are con- sequently often no longer exposed to ETS at home. Divorce or death of the spouse may also terminate home ETS exposure, the latter more commonly among women. On the other hand, more never-smoking men are married to currently or recently smoking women. As an example, a recent exposure assessment study in Stockholm that included 182 subjects managed to re- cruit only 10 men and 11 women with home exposure?° Our data also indicate that the average amount smoked at home among currently smoking men married to nev- er-smoking women has decreased over the last decades, whereas the opposite is true for women. These factors can increase exposure misclassification, especially among women. For exposure variables that fail to address such exposure heterogeneity, any effect of spousal ETS exposure may be difficult to discern. Our study shows a clearer lung cancer effect for vari- ables measuring ETS exposure in the work place. Indi- vidual monitoring also indicates that in Sweden today, work and home ETS exposure are of similar intensity for those exposed, but work exposure is more common.30 If recent ETS exposures are biologically most relevant, less misclassification may result with variables of work expo- sure. This possibility offers one explanation for the un- expectedly higher RRs observed for work exposure than for spousal ETS exposure among women in this and several other studies.3-5,9,31-33 Our data indicate higher RRs for measures of spousal ETS exposure in subjects below 70 years of age. This age modification may reflect less misclassification owing ei- ther to more accurate reporting or to better correspon- dence with biologically relevant recent exposures (the average time since last exposure was 14.5 years among control subjects younger than 70 years, and 21.9 years among those age 70 years or older). It could also reflect a true age interaction. Our data also showed stronger effects for exposure to spousal ETS among men than among women. Although the confidence intervals are wide, the finding may reflect that ETS exposures among men are recent and of high intensity. Our study provided an opportunity to investigate ex- posure timing, owing to the variability in exposure pat- terns. The analyses indicate that exposure continuing into the recent 10- to 15-year period may be most important, consistent with the decreased excess risk demonstrated after stopping active smoking.is-17 Previously, only three studies have analyzed lung can- cer risk and time since last exposure to ETS. One early Greek case-control study34,35 used 0-4 years ago for "current" exposure and 5 or more years ago for "former"

306 NYBERG ET AL Epidemiology May 1998, Volume 9 Number 3 TABLE 4. Relative Risk of Lung Cancer Associated with Reported Exposure to Major Sources of Environmental Tobacco Smoke,* Men and Women Combined I Variable Cases Controls RRt 95% CIt Years since last exposure from major source* Never any major exposure$ 16 36 1 >20 20 51 1.04 0.46-237 11-20 30 53 1.41 0.65 3.06 3-10 26 52 1.12 0.50-2.50 0-2 32 43 2.12 0.91-4.92 Total duration of major exposure (years)§ Never any major exposure$ 16 36 1 1-5 9 24 0.98 0.35-2.75 6-20 32 65 1.23 0.56-2.70 21-34 34 56 1.51 0.69-3.32 34-45 21 31 1.60 0.68-3.75 >45 12 23 1.10 0.42-2.88 Highest duration category for either of the two major exposure sources1l Never any major exposure$ 16 36 1 Lowest quartile of both 17 36 1.20 0.49-2.93 Max second quartile, in one or both 21 45 1.06 0.45-2.48 Max third quartile, in one or both 29 53 1.32 0.59-2.93 Max 75th-90th percentile, in one or both 19 40 1.20 0.51-2.80 Above 90th percentile in one or both 22 25 1.84 0.77-4-37 Hour-years'l of major exposure§ Never any major exposure$ 16 36 1 0<...:5.5 HY 9 23 0.98 0.35 2.73 5<...s22 HY 27 57 1.15 0.52-2.56 22<...s65 HY 30 58 1.31 0.60-2.85 65<...s120 HY 11 35 0.72 0.28-1.86 > 120 HY 25 24 2.50 1.05-5.92 Highest hour-yearslf category for either of the two major exposure sources1l Never any major exposuret 16 36 1 Lowest quartile of both 11 28 0.96 037-2.50 Max second quartile, in one or both 30 53 1.33 0.61-2.93 Max third quartile, in one or both 17 55 0.72 0.31-1.68 Max 75th-90th percentile, in one or both 22 37 1.58 0.68-3.66 Above 90th percentile in one or both 28 26 2.52 1.08-5.85 Total number of cases or controls varies slightly between analyses owing to missing data for some subjects for the respective exposure variables. * Exposure from spouse or work. t Relative risk (RR) and 95% confidence interval (CI) estimated by odds ratios obtained in multiple logistic regression. Adjusted for matching variables gender, age, and catchment area, as well as occasional smoking, vegetable consumption, degree of urban residence, and years of exposure to risk occupations. # Referent category. § Category limits among exposed are the 25th, 50th, 75th, and 90th percentiles of the distribution among all controls. IF One "hour-year" is 365 hours, or the equivalent of I hour per day for I year. 11 Calculated from the variables for environmental exposure from work or spouse separately, based on quantiles of the exposed controls within each variable. Subjects are classified into the category corresponding to the highest category for either of the two single exposure quantile variables. exposure, finding a crude RR of 1.9 in both groups. Among exposed controls, 69.8% had "current" exposure, and even most subjects with "former" exposure were probably exposed within 5-10 years. In a Japanese case- control study,36 "current" exposure was within 0-9 years (49.4% of exposed controls), and "former" exposure was 10 years or more ago. The RRs were 1.8 (95% CI = 1.0 3.2) and 1.3 (95% CI = 0.9-2.4), respectively. In an early Japanese cohort study37 on nonsmoking women, 91.1% of exposed person-years were for subjects with current ETS exposure at the start of follow-up. From reported changes in male smoking habits,37 an estimate is that 70-80% of women "currently exposed" at base- line had ETS exposure from a regularly smoking spouse in the 10 years before the end of follow-up, compared with 30-50% among women "non-currently exposed" at baseline. The RRs were 1.45 (95% CI = 1.04-2.02) in the "current at baseline/70-80% recently" exposure cat- egory and 1.36 (95% CI = 0.85-2.18) in the "non- current at baseline/30-50% recently" category. The pathology review results are similar to those of another large study of ETS and lung cancer38 and make bias through misdiagnosis unlikely. Carcinoid lung tumors were included, although they are often considered not related to smoking.39 Nevertheless, many reports also indicate a relation,4°-44 but it seems weaker than for squamous and small cell cancers. ETS as a risk factor for carcinoids has not been evaluated. Furthermore, differences between mainstream and sidestream smoke could give different risk profiles for active and passive smoking, as proposed for adenocar- cinomas.45-47 Among our never-smokers, 12 cases and 24 controls had smoked 20-408 packs on an occasional basis (0.05- 1.12 pack-years), and of these, only 11 had smoked in the last 10 years. These "occasional smokers" were in- cluded in the analyses. We explored possible confound- ing of the ETS relation using categorical and continuous variables for occasional smoking, with no such evidence. Confounding from other risk factors including urban .::.~<

Epidemiology May 1998, Volume 9 Number 3 TABLE 5. Relative Risks* of Lung Cancer with 95% Confidence Intetvaisi' for Exposure to Major Sources of Environmental Tobacco Smoke (Spouse and Work), Men and Women Combined. In Both Analyses, Subjects Unex- posed to Either Source Constitute the Reference Group Time since Last Cumulative Exposure (Hour-Years)t Ex osure (Years) :522.9,t. >22.9 >15$ 3-15 0-2 0.83 (0.34-2.06) 1.30 (0.52-3.25) 2.25 (0.67-7.53) 1.11 (0.45-2.76) 1.32 (0.61-2.87) 1.87 (0.77-4.63) Total Duration of Exposure (Years) 1-20$ >20$ >15# 1.13 (0.49-2.59) 0.98 (0.37-2.57) 3-15 1.14 (0.43-3.02) 1.39 (0.64-2.99) 0-2 2.27 (0.67-7.66) 2.09 (0.86-5.09) * Estimated by odds ratios obtained in multiple logistic regression. Adjusted for matching variables gender, age and catchmenc area, as well as occasional smok- ing, vegetable consumption, degree of urban residence, and years of exposure to risk occupations. t One "hour-year" is 365 hours, or the equivalent of 1 hour per day for 1 year. t Cutoff represents 50th percentile among controls. residence, occupational lung carcinogens, and diet also did not appear to explain the results. Validation of the subject's ever-smoking status by interview with next-of-kin showed very low misclassifi- cation rates, and the discrepant reports reflected more differing interpretations than true differences in the re- ported smoking habits. Any effect on ETS estimates by these levels of misclassification would be negligible.lal,2a In conclusion, we found evidence of an effect of adulthood ETS exposure on lung cancer risk, particularly for occupational exposure. Exposure patterns for spousal ETS were different between genders, and we found clearer excess risks among men, who had higher expo- sures more recently. Combining ETS exposure from work and spouse provided the strongest evidence of elevated risks in high-exposure groups and of recent exposure as a major determinant of lung cancer risk from ETS exposure. Our findings are consistent with previous Swedish studies48,49 and the continuing changes in smok- ing habits observed in Sweden. They highlight the im- portance of considering timing of exposure and exposuce from different sources when studying lung cancer risk and ETS exposure. Acknowledgments We are grateful for the cooperation of the departments of respiratory disease and pathology of the participating hospitals, and of all staff who helped make this study possible. We also thank Klaus-Michael MUller of the Institute for Pathol- ogy, Berufsgenossenschaftliche Kliniken Bergmannsheil in Bochum, for careful histopathologic review; Carin Cavalli-BjSrkman for dedicated field and data collection work, as well as Inger Isaksson, Eva-Brict Gustavsson, Gun Johnsson, Mette Lindevall and others at the Karolinska Institute for valuable contributions; and Martin Kmus and Gilles Ferro at the International Agency for Research on Cancer for programming support. ETS AND LUNG CANCER 307 References 1. U.S. Environmental Protection Agency. Respiratory Health Effects of Pas- sive Smoking: Lung Cancer and Other Disorders. Report No. EPA/600:6- 90/006F. Washington DC: U.S. Environmental Protection Agency, 1992. 2. Stockwell HG, Goldman AL, Lyman GH, Noss CI, Armstrong AW, Pinkham PA, Candelora EC, Brusa MR. Environmental tobacco smoke and lung cancer risk in nonsmoking women. J Natl Cancer Inst 1992;84:1417- 1422. 3. Brownson RC, Alavanja MC, Hock ET, Loy TS. Passive smoking and lung cancer in nonsmoking women. Am J Public Health 1992;82:1525-1530. 4. Fontham ET, Correa P, Reynolds P, Wu-Williams A, Buffler PA, Greenberg RS, Chen VW, Alterman T, Boyd P, Austin DF, Liff J, Greenberg SD. Environmental tobacco smoke and lung cancer in nonsmoking women: a multicenter study. JAMA 1994;271:1752-1759. 5. Kabat GC, Stellman SD, Wynder EL. Relation between exposure to envi- ronmental tobacco smoke and lung cancer in lifetime nonsmokers. Am J Epidemiol 1995;142:141-148. 6. Cardenas VM, Thun MJ, Austin H, Lally CA, Clark WS, Greenberg RS, Heath CW. Environmental tobacco smoke and lung cancer mortality in the American Cancer Society's Cancer Prevention Study 11. Cancer Causes Control 1997;8:57-64. 7. Ko YC, Lee CH, Chen MJ, Huang CC, Chang WY, Lin HJ, Wang HZ, Chang PY. Risk factors for primary lung cancer among non-smoking women in Taiwan. Inc J Epidemiol 1997;26:24-31. 8. Liu Q, Sasco AJ, Riboli E, Hu MX. Indoor air pollution and lung cancer in Guangzhou, People's Republic of China. Am J Epidemiol 1993;137:145- 154. 9. Schwartz AG, Yang P, Swanson GM. Familial risk of lung cancer among nonsmokers and their relatives. Am J Epidemiol 1996;144:554-562. 10. Wu-Willianss AH, Samet JM. Environmental tobacco smoke: exposure- response relationships in epidemiologic studies. Risk Anal 1990;10:39-48. 11. Dockery DW, Trichopoulos D. Risk of lung cancer from environmental exposures to tobacco smoke (Review). Cancer Causes Control 1997;8:333- 345. 12. Correa P, Pickle LW, Fontham E, Lin Y, Haenszel W. Passive smoking and lung cancer. Lancet 1983;2:595-597. 13. Janerich DT, Thompson WD, Varela LR, Greenwald P, Chorost S, Tucci C, Zaman MB, Melamed MR, Kiely M, McKneaily MF. Lung cancer and exposure to tobacco smoke in the household. N Engl J Med 1990;323:632- 636. 14. Tobacco Smoking: IARC Monographs on the Evaluation of the Carcino- genic Risk of Chemicals to Humans. vol. 38. Lyon: International Agency for Research on Cancer, 1997. 15. Becher H, Jackel KH, Timm J, Wichmann HE, Drescher K. Smoking cessation and nonsmoking intervals: effect of different smoking pattems on lung cancer risk. Cancer Causes Control 1991;2:381-387. 16. Lubin JH, Blot WJ, Berrino F, Flamant R, Gillis CR, Kunze M, Schmahl D, Visco G. Modifying risk of developing lung cancer by changing habits of cigarette smoking. BMJ 1984;288:1953-1956. 17. Lubin JH. Modifying risk of developing lung cancer by changing habits of cigarette smoking (Letter). BMJ 1984;289:921. 18. Matanoski 0, Kanchanaraksa S, Lantry D, Chang Y. Characteristics of nonsmoking women in NHANES I and NHANES 1 epidemiologic fol- low-up study with exposure to spouses who smoke. Am J Epidemiol 1995; 142:149-157. 19. Thomton A, Lee P, Fry J. Differences between smokers, ex-smokers, passive smokers and non-smokers. J Clin Epidemiol 1994;47:1143-1162. 20. Lee PN. Misclassification of Smoking Habits and Passive Smoking: A Review of the Evidence: International Archives of Occupational and Envi- ronmental Health Supplement. Berlin: Springer-Verlag, 1988. 21. Nyberg F, Isaksson 1, Harris JR, Pershagen G. Misclassification of smoking status and lung cancer risk from environmental tobacco smoke in never- smokers. Epidemiology 1997;8:304-309. 22. Lee PN. Environmental Tobacco Smoke and Mortality. Basel: Karger, 1992. 23. World Health Organization. The World Health Organization Histological Typing of Lung Tumours. 2nd ed. Am J Clin Pathol 1982;77:123-136. 24. Riboli E, Preston-Martin S, Saracci R, Haley NJ, Trichopoulos D, Becher H, Burch JD, Fontham ET. Gao YT, Jindal SK, Koo LC, Le Marchand L, Segnan N, Shimizu H, Stanta G, Wu-Williams AH, Zatonski W. Exposure of nonsmoking women to environmental tobacco smoke: a 10-country collaborative study. Cancer Causes Control 1990;1:243-252. 25. Boffetta P, Kogevinas M, Simonato 1„ Wilboum J, Saracci R. Current perspectives on occupational cancer. Int J Occup Environ Health 1996;1: 315-325. 26. World Health Organization. International Standard Code for Occupations 1968. Geneva: World Health Organization, 1968. 27. World Health Organization. International Standard Industrial Code 1971. Geneva: World Health Organization, 1971. 28. Nyberg F, Agudo A, Boffetta P, Forces C, Gonz3lez CA. Pershagen G. A European validation study of smoking and environmental tobacco smoke ._....- .~r

308 NYBERG ET AL exposure in non-smoking lung cancer cases and controls. Cancer Causes Control 1998;9:173-182. 29. Nicolaides-Bouman A, Wald N, Forey B, Lee P. International Smoking Statistics: a Collection of Historical Data from 22 Economically Developed Countries. New York: Oxford University Press, 1993. 30. Phillips K, Bentley MC, Howard DA, Alvan G. Assessment of air quality in Stockholm by personal monitoring of nonsmokers for respirable suspended particles and environmental tobacco smoke. Scand J Work Environ Health 1996;22:1-24. 31. Shimhu H, Morishita M, Miruno K, Masuda T, Ogura Y, Santo M, Nish- imura M, Kunishima K, Karasawa K, Nishiwaki K, Yamamoto M, Hisamichi S, Tominaga S. A case-control study of lung cancer in nonsmoking women. Tohoku J Exp Med 1988;154389-397. 32. Wu AH, Henderson BE, Pike MC, Yu MC. Smoking and other risk factors for lung cancer in women. J Natl Cancer Inst 1985;74:747-751. 33. Wu-Williams AH, Dai XD, Blot W, Xu ZY, Sun XW, Xiao HP, Stone BJ, Yu SF, Feng YP, Ershow AG, Sun J, Fraumeni JF Jr, Henderson BE. Lung cancer among women in north-east China. Br J Cancer 1990;62:982-987. 34. Trichopoulos D, Kalandidi A, Sparros L, MacMahon B. Lung cancer and passive smoking. Int J Cancer 1981;27:1-4. 35. Trichopoulos D, Kalandidi A, Sparros L. Lung cancer and passive smoking: conclusion of Greek study (Letter). Lancet 1983;2:677-678. 36. Akiba S, Kato H, Blot WJ. Passive smoking and lung cancer among Japanese women. Cancer Res 1986;46:4804-4807. 37. Hirayama T. Life-style and mortality: a large-scale census-based cohort study in Japan. In: Wahrendorf J, ed. Contributions to Epidemiology and Biosta- tistics. vol. 6. Basel: Karger, 1990. 38. Fontham ET, Correa P, WuWilliams A, Reynolds P, Greenberg RS, Buffler PA, Chen VW, Boyd P, Alterman T, Austin DF, Liff J. Lung cancer in Epidemiology May 1998, Volume 9 Number 3 nonsmoking women: a multicenter case-control study. Cancer Epidemiol Biomarkers Prev 1991;1:35-43. 39. Davila DO, Dunn WF, Tazelaar HD, Pairolero PC. Bronchial carcinoid tumors. Mayo Clin Proc 1993;68:795-803. 40. Froudarakis M, Fournel P, Burgard G, Bouros D, Boucheron S, Siafakas NM, Emonot A. Bronchial carcinoids: a review of 22 cases. Oncology 1996;53: 153-158. 41. Grote TH, Macon WR, Davis B, Greco FA, Johnson DH. Atypical carcinoid of the lung: a distinct clinicopathologic entity. Chest 1988;93370-375. 42. Harpole DH Jr, Feldman JM, Buchanan S, Young WG, Wolfe WG. Bron- chial carcinoid tumors: a retrospective analysis of 126 patients. Ann Thorac Surg 1992;54:50-55. 43. Mills SE, Cooper PH, Walker AN, Kron IL Atypical carcinoid tumor of the lung: a clinicopathologic study of 17 cases. Am J Surg Pathol 1982;6:643-654. 44. Valli M, Fabris GA, Dewar A, Hornall D, Sheppard MN. Atypical carcinoid tumour of the lung: a study of 33 cases with prognostic features. Hisropa- thology 1994;24:363-369. 45. Wynder EL, Goodman MT. Smoking and lung cancer: some unresolved issues. Epidemiol Rev 1983;5:177-207. 46. Wynder E., Kabat GC. The effect of low-yield cigarette smoking on lung cancer risk. Cancer 1988;62:1223-1230. 47. Charloux A, Quoix E, Wolkove N, Small D, Pauli G, Kreisman H. The increasing incidence of lung adenocarcinoma: reality or artefact? A review of the epidemiology of lung adenocarcinoma. Int J Epidemiol 1997;26:14-23. 48. Pershagen 0, Hrubec 2., Svensson C. Passive smoking and lung cancer in Swedish women. Am J Epidemiol 1987;125:17-24. 49. Svensson C, Pershagen G, Klominek J. Smoking and passive smoking in relation to lung cancer in women. Acta Oncol 1989;28:623-629.