Tobacco Documents Online

! I I I I I I I I I I I ENV[RONMENTAL TOBACCO SMOKE AND LUNG CANCER MORTALITY IN TI~ AMERICAN CANGER SOCIETY'S CANCER PREVENTION STUDY [[ by Victor M. Cardenas, M.D., National Autonomous University of Mexico, 1978 Epidemiologist, School of Public Health of Mexico, 1981 M.P.H., Emory University, 1990 Adviser: Harland D. Austin, D.S.c. A dissertation submitted to the Faculty of the Graduate School of Emory University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Division of Epidemiology 1994 i

Dedicated to dearest Carmen and Angela whose love has been a constant source of support, always! 0 ,CO 0 0 I I ! I i I I I I I I I I I I

I I I i" I I I i I I i I I I I I I I Acknowledgments Without the support, encouragement and help of many individuals the completion of this dissertation and academic degree would have been an impossible task. I am indebted to my mentors, Drs. John Boring, Jonathan Lift, Ira Longini, Ray Greenberg, Ralph Coates, Harland Austin and Dana Flanders for their patient instruction, the learning opportunities they provided me, and for their insistence on excellence in epidemiologic research. The vision and encouragement of Dr. Ray Greenberg was largely instrumental in my involvement in this program. Our program advisor, Dollie Daniels, deserves special mention for her support throughout the past four years has been much appreciated. I also wish to acknowledge the excellent proofreading of Ms. Daniels and Dr. Louis Escobedo of several versions of this manuscript. I am especially grateful to each member of my doctoral dissertation committee, Drs. Harland Austin, Michael Thun, Clark Heath, Scott Clark, and Ray Greenberg for their support, patience and encouragement throughout this experience. I have been especially fortunate to have Harland Austin as advisor in the area of cancer epidemiology and epidemiologic methods. He was always there to omaide me through the different stages of this study. At Emory Universfiy I have had the privilege to use the American Cancer Society Cancer Prevention Study 1I data base and facilities for my research. It has been a pleasure to work and receive advice on a daily basis from Drs. Michael Thun, and Clark Heath. The programming guidance of Mohan Namboodiri first, and then brilliantly embraced by Cathy Lally and Heidi Miracle, was decisive. I also acknowlpdge the advice of Dena Myers, Dr. Eugenia Calle, and Diane Terrell for their valuable suggestions throughout the study conduct.

To develop a definition of different categories of jobs/trades with potential occupational exposure to asbestos, I was fortunate to count on with the advice of an expert in the field such as Dr. Howard Frurnkin, and I also want to express my gratitude to him. I thank Dr. Christina Park, from the National Center for Health Statistics, for providing unpublished data from the 1988 NHIS-OH supplement on the extent of ETS exposure. I thank the following persons, principal investigators of the National Cancer Institute SEER program, who promptly replied to our inquiry to confirm the diagnoses of lung cancer among nonsmoking CPS II participants who resided in the Cancer Registries areas: Mrs. Kathleen M. McKeen; Director of the State Health Registry of Iowa; Mr. J.T. Flannery, Director of the Connecticut Tumor Registry; Dr. David B. Thomas, from the Cancer Surveillance System of Western Washington; Dr. Dee West, Director of the San Francisco Bay Area Cancer Registry; Dr. Charles R. Key, Medical Director of the New Mexico Tumor Registry; to my professor, Dr. Jonathan M. Lift, Director of the Georgia Center for Cancer Statistics; Dr. G. Marie Swanson Director, and Sherry Enoex, Data Manager from the Michigan Cancer Foundation; Dr. L.N. Kolonel, Director and Ruth D. Mertz, administrator from the Cancer Research Center of Hawaii; and to Dr. W.P. McWhorter Director of the Utah Cancer Registry. Finally, I acknowledge the generous support of the W.K. Kellogg Foundation and Emory University for providing me with scholarships to complete my studies. 0 I I I I i I I I I I I I i I

I I I I- I I I I I I I I I I I I I I NOTICE TO BORROWERS - Unpublished theses deposited in the Emory University Library must be used only in accordance with the stipulations prescribed by the author in the preceding statement. The author of this dissertation is: Vfctor Manuel C~denas-Ayala Transversal 13-124 Int 4-103 Santa F6 de Bogot~ Colombia The director of this dissertation is: Harland D. Austin 4870 Forestglade Ct. Stone Mr. GA 30087 Users of this dissertation not reg~flarly enrolled as students at Emory University are required to attest acceptance of the preceding stipulations by signing below. Libraries borrowing this dissertation for the use of their patrons are required to see that each user records here the information requested Name of user Address Date Type of Use (Examination only or copying)

I I I ENVIRONMENTAL TOBACCO SMOKE AND LUNG CANCER MORTALITY IN THE AMERICAN CANCER SOCIETY'S CANCER PREVENTION STUDY II by Victor M. Cardenas, M.D., M.P.H. Adviser: Harland Austin, D. Sc. Division of Epidemiology Approved for the Division Adviser Date Accepted: Date 0 / I I I I I I I I I I I I

I I I I I I I I I I I I I I The dissertation of Victor M. Cardenas is approved. Committee Members Harland D. Austin, D. Sc., Chair Michael J. Thun, M.D., M. Sc. Clark W. Heath Jr., M.D. Scott W. Clark, Ph.D. Raymond S. Greenberg, H., Ph.D.

vi I I i I I ENVIRONMENTAL TOBACCO SMOKE AND LUNG CANCER MORTALITY IN THE AMERICAN CANCER SOCIATY'S CANCER PREVENTION STUDY II by Victor M. Cardenas, M.D., National Autonomous University of Mexico, 1978 Epidemiologist, School of Public Health of Mexico, 1981 M.P.H, Emory University, 1990 Adviser: Harland D. Austin, D. Sc., An Abstract of A dissertation submitted to the Faculty of the Graduat~ School of Emory University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Division of Epidemiology 1994 I I I I I I I i I I

I I I I I I I I I I I I I I tt~aoie ot ~ontents Chapter 1 Introduction 1 1.1. Backgound 1.2. Specific Aim 1.3. Hypotheses 1.4. Definitions of ETS, Lung Cancer, and Analytical Cohorts ETS in General ETS Exposure in CPS II Lung Cancer in General Ltmg Cancer in CPS II Self-reported and Spousal Exposure to ETS 1.5. Significance and Relevance of the Study 1.6. Extent ofETS E.xposure 1.7 Organization of Thesis Chapter 2 Literature Review and Metanalysis 2.1. Descriptive Lung Cancer Epidemiology 2.2 Smoking and Lung Cancer 2.3. Biological Plausibility 2.4 Epidemiological Evidence 2.5 Collateral Evidence 2.5 Collateral Evidence 2.6 Controversy 2.7 Other Risk Factors 2.8 Lung Cancer Classification 2.9 Measurement of ETS Exposure Chapter 3 Methods 3.1 The Cancer Prevention Study II Study Population The Cancer Prevention Study II a) Recruitment b) Follow-up c) Baseline Questionnaire 3.2. Published Results from CPS II 3.3 Main Design Features 3.4. Sub Cohort of Interest Definition and Source Population 3.5. Eligibility a) Inclusion Criteria b) Exclusion Criteria 3.6. Variables Status Page l 2 '3 3 3 4 4 5 5 5 7 9 11 12 13 14 16 19 20 24 25 26 28 28 28 29 29 o 3O 30 31 31 31 32 33 33 0 I 0

Mam Outcome Exposure Variables i.Main Exposure Variable (Self-reported ETS) ii. Spousal Smoking Habits Covariates Denominator Information 3.7. Validation of ETS exposure data 3.8. Exposure Criteria Used In Analysis i. Self-reported ETS Exposure ii. Spousal ETS Exposure 3.9. Validation of Information from Death Certificates 3. I0. Statistical Analysis Outline of the Analytic Approach Exploratory Data Analysis Simple and Stratified Analysis Multivariate Analysis 3.11. Sample Size and Power Considerations Chapter 4 Comparisons of Demographics and Smoking l:[abits in the US, CPS II, and the Study Populations Rationale Race Gender and Age Structure Occupation Schooling Marital Status Residence Smoking Habits Chapter 5 Validity and Completeness of the Information On the Outcome Variable i. Follow-up Procedures ii. Results of Validation Study of Death Certificate Diagnoses of Lung Cancer Main Exposure Variables i. Self Reported ETS Exposure in CPS II and NI-IIS ii. CPS II Self-reported Exposure to ETS and Spousal Smoking Habits Chapter 6 Descriptive Statistics of Exposure Variables Frequency of Self-reported and Spousal ETS Correlates of ETS exposure 36 37 37 37 38 40 40 43 43 43 43 44 44 47 47 51 52 54 54 54 56 58 59 60 60 63 63 63 64 64 71 75 76 I I l I i I ! I I ! I I I I I i i I

I I I I- I I I I I I I I I i I I I I Caapter 7 Main Results 7.1 Deaths from Lung Cancer and H.istological Dam in Death Certificates 7.2 Potential confounders age, gender, schooling, race, prexisting lung disease, occupational exposure to lung carcinogens, consumption of foods containing carotenoids, and fat as nutrient index. A~ Gender Race Schooling ~ Asbestos Prexisting chronic lung disease Diet Other 7.3.a. Self reported ETS 7.3.b. Dose-response of S,elf reported ETS 7.3.c. Spousal ETS 7.4 Joint effects of'ETS and asbestos exposure 7.5 Leading causes of death in the cohort Chapter 8 Discussion and Conclusions ... 8.1 Consistency 8.2 Study Power 8.3 Misclassification 8.4 Confounding 8.5 Misclassification of smoking status 8.6 Causal Inference : 8.7 Conclusions References Appendices 85 86 86 87 88 89 90 92 94 94 97 97 107 123 128 129 133 134 138 139 141 143 146 158 o 0~ 0~ o

List of figures Page I. RR of lung cancer for ETS from 36 epidemiologic studies ......................... 18 2. Funnel plot of RR fro ETS from 36 studies .......................................... 19 3. Hypothetical cohort study: effect of misclassification of active smoking on lung cancer risk ............................... 22 4. Exclusions for dose-response analyses ............................................... 36 5. Outline of analytical approach .......................................................... 46 6. Self-reported ETS of spouses and cigarette smoked by their husbands ........... 73 7. Rates of lung cancer among nonsmokers by age ..................................... 86 8. Rates of lung cancer among nonsmokers by ETS exposure ........................ 104 9. Rates of lung cancer among nonsmokers by ETS from spousal smoking ...................................................................................... 110 10. Log-log (S) curves for ETS .......................................................... 126 List of Tables ............................................................................ Page 1. Extent of ETS Exposure in the USA ................................................ 9 2 Metanalysis from 36 published studies ............................................. 17 3. Smoking habits at time of enrollment ............................................... 34 4 Analytic cohorts ........................................................................ 35 5 Comparison of age- men US,CPS 17, and analytic cohorts ..................... 55 6. Comparison of age : women US, CPS 17, and analytic cohorts ................. 56 7. Comparison of occupations of employed US, CPS II, and analytic cohorts .............................................................................. 57 8. Comparison of college graduates US,CPS II, and analytic cohorts ............ 58 9. Comparison of married whites in the US, CPS lr, and analytic cohorts ................................................. : ............................ 59 10. Comparison residence US, CPS II participants .................................... 61 11. Age-adjusted prevalence smoking, CPS 1-[ and NI-~S-83 ......................... 62 12. CPS 17 on reported hours of ETS exposure at different settings .................. 65 13. Patterns of answers given by CPS II participants to seE-assessment ............ 66 14. Nonsmokers reportedly exposed to ETS at home CPS 17 and NHIS ........... 69 15. Characteristics of CPS II nonsmokers in analytic cohort for self- reported ETS analysis by completeness of information on ETS ....................... 70 16. Comparison of self-reported exposure to ETS at home by CPS rr nonsmoking women and the number of current smokers among cohabitees, and spouses .................................................................... 71 17. Distribution of reported hours of exposure to ETS at home by nonsmoking women, according to number of cigarettes smoked by their husbands ............................................................................... 74 18. Hours of exposure to ETS reported by nonsmoking CPS participants at differente settings, 1982 ................................................................. 76 I I I I I I ! I I I i I

I I I I I I I I I I I I I I I I I 19. t'ropomon of nonsmo~tng ~.PS Lt pamcipants m analytic cohorts who reported any ETS exposure by age at interview, 1982 ........................... 78 20. Age of nonsmoking participants by ETS from spousal smoking ................ 78 21. Characteristics of nonsmoking participants according to self-reported ETS .............................................................................. 80 22. Characteristics of nonsmoking participants according to ETS from spousal smoking ..........~ ............................................................ 81 23. Intensity and duration of spousal cigarette smoking by spouses ................. 84 24. Distribution of hystological types in 115 deaths from lung cancer .............. 85 25. Rates of Lung Cancer among Nonsmokers by Age and Gender ................. 87 26. Rates of Lung Cancer among Nonsmokers by Age and Race .................... 88 27. Rates of Lung Cancer among Nonsmokers by Age and Schooling .............. 89 28. Rates of Lung Cancer amohg Nonsmokers by Asbestos exposure at work .............................................................................. 90 29. Rates of Lung Cancer among Nonsmokers by years of exposure to asbestos .............................................................................. 93 30. Rates of Lung Cancer among Nonsmokers by non ETS risk factors ............ 96 3 I. Lung cancer mortality and risk estimates by self-reported exposure to ETS (any versus none) .........!. ........................................................... 99 32. Lung cancer mortality and risk estimates by self-reported exposure to ETS (by tertiles) ............................................................................. I00 33. Multivariate risk estimates of lung cancer by self-reported exposure to ETS .............................................................................. I06 34. Lung cancer rate ratios among nonsmokers by smoking status of their husbands (current cigarettds vs. never) ................................................... 109 35. Lung cancer rate ratios among nonsmoking wives by smoking status of their husbands (current pipe/cigar vs. never) ......................................... 1 I2 36. Lung cancer multivariate rate ratios by type of smoking of spouses ............. 113 37. Lung cancer multivariate rate ratios by amount of cigarette smoking ............ 115 38. Lung cancer multivariate rote ratios by amount of cigarette smoking (among spouses married once)! ........................................................... 117 39. Lung Cancer multivariate rate ratios by time in marriage with cigarette smoking spouses ........................................................................... 40. Lung Cancer multivariate rote ratios by pack-years of cigarette smoking spouses ............................................. ., .............................. 122 41. Rate ratios for lung cancer by ETS and asbestos exposure ....................... 124 42. Other causes of death .................................................................. 128 43. Power calculations ..................................................................... 134 44. Corrected RR by degree of misclassification of ETS exposure ................... 135 45. Rate ratios for lung cancer by ETS by approach to missing data ................. 136 46. ETS exposure by education ........................................................... 137 0 O~ O~ 0

i III I April 5, 1954. 1978 1977 1979-1980 1981 1982-1985 1985-1987 1987-1989 1989-1990 Vita Born, Mexico, D.F. M.D. National University, Mexico City Medical doctor at the EI DuPont de Nemours manufacturing plant in Mexico City. Researcher. Directorate of Occupational Health, Ministry of Labor. Degree of Epidemiologist School of Public Health of Mexico. Epidemiologist, Directorate of Epidemiology, Ministry of Health. Degree of Epidcmiologist Mexican Field Epidcmiology Training Program. Ministry of Health, School of Public Health of Mexico, and Centers for Disease Control. Associate P, csearchcr and Professor Center for Public Health Research and School of Public Health of Mexico. National Institute of Public Health. Master of Public Health Emory University. 0 O~ 0 I I I i I I I I I I I I / I

I I I I- I I I I I I I I I I I I I I ! iv Fublicadons 1) C~denas-Ayala, VM.,I Montesano-Castellanos R, Morales-Marfn, R Villegas-Espinosa, A mad Flores-Salinas, E: Prevalence and incidence estimates of paralytic poliomyelitis by means of lanmess surveys in Tecom~n, Colima. Sal Ptib M~x 1985, 27 (2): 52-59. (Published in Spanish). 2) C~denas-Ayala, VM, Koopman James S, Guerrero-Reyes P: Tuberculosis mortality by birth cohorts, b, lexico. Bol. Mensual Epiderniologia del Sector Salud 1986; I: 137-141. (Pdblished in Spanish). 3) Dfaz-Ortega JL, Valdespino-G6mez JL, ZArate-Aquino ML, Ruiz-Mams C, C~denas-Ayala, VM and Solache G: Sero-immune response to measles vaccine in children 8 -18 months of age. Bol M~d Hosp Infant (M6x) 1986; 43: 526-532. (Published in Spa~.ish). 4) C~denas-Ayala, VM: Factors contributing to ARI mortality. ARI News 1987; 1 (7): 5. (Published in English). Also in Noticias sobre IRA: 1987, p.19-20. (Published in Spanish). 5) Ruiz-Mams C, .K.oopman)James S, Cgrdenas-Ayala VM, Herrera-Bastos E and I-Iinojosa-Martmez M: 'Diarrhoeal disease after the 1985 earthquakes in Mexico City. Sal Ptib Mdx 1987; 29: 399-405. (Published in Spanish). 6) C~denas-Ayala VM, Vilchis-Lic6n H, Stetler Harrison C, Koopman James S, Cabrera-Coello L, Valdespino-G6mez JL, Vega G and Muro Amador M: Risk factors for wild poliov~rus transmission in Sinaloa, Mexico, 1984-1986. Bol Of Sanit Panam 1988; 105(1): 1-12. English version of the same paper in: PAHO Bull 1988; 22 (3): 227-239. Also reprinted in Infectologia 1988; 8 (11): 571-580. 7) C~rdenas-Ayala VM, Bemal-P6rez J, Stetler Harrison C, Guerrero-Reyes P: Tuberculin skin surveys in Guerrero and new estimates for tuberculous infection frequency in Mexico. Sal INb Mdx 1988; 30: 152-157. (published in Spanish). 8) Pefia Corona MP, C~denas-Ayala VM, Ruiz Mares C, Stetler HC, L6pez Ortiz O, Ibarra Rosales J, Sapiafn L6pez LA and VillafLn FS: Whooping cough outbreak in San Bartotom6 Quialana, 1988. Report of the field investigation. Bol. Mensual Epidemiologia del Sector Salud 1989; 4 (8): 112- 118. (Published in Spanish). 9) Cgtrdenas-Ayala VM, Ruiz-Matus C, Montesano-Castellanos R, Cabrera- Coello L, Garrido-Lozada T, OIvera-Castillo, R and Diaz-Ortega JL: Cost- benefit analysis of measles vaccination program in Mexico. Sal Ptib M6x 1989; 31 (6): 735-44. (Published in Spanish). 0

v it)) (2arctenas-Ayaia VM anct bancttez, ~2: 5octal impact or measles vaccination programs in Mexico. Bol. M6d Hosp IMantil (M6xico) 1990; 47 (7) • 500-505. (Published in Spanish). i 1) Stem C, Tolbert K, Nufiez R_M, Goodwin M, Cg.rdenas VM: Changing conditions of child survival in M6xico and strategies for the furore. S~1 M6x 1990; 32 (5): 532-542. (Published in Spanish). 12) Garrido F, Borgues GG, C~irttenas VM, Ibarra JM, Borgues GG: Po.stneonatal mortality from diarrheal diseases: a case-control study. S~fl Mex 1990; 32 (3): 261:-268. (Published in Spanish). 13) C~denas-Ayala VM, Koopman James S, Garrido-Latorre F, Bazxia Luis F and Ibarra-Rosales, Alag6n J and Stetler, Harrison C.: Protective effect of antibiotics on mortality from acute respiratory infections. PAHO Bull 1992 26 (2): 109-120. Also published in Spanish in Bol Of Sanit Panam 1993; 114 (3); 202-212. 14) Flanders DW, C~denas VM, Austin H: Confounding by time since hire in internal comparisons of cumulative exposure in occupational cohort studies. Epidemiology 1992; 4:336-341. 15) Cgrdenas VM, Saad C, Varona M, and Linero M: Waterborne cholera in Riohacha, 1992. PAHO Bull 1993; 27 (4): 313-336. 16) C,'frdenas VM, Smith TE, Nufiez RM, Brogan DN, Gatica N, Ibarra JM, Galv~fn S, Flores ME, and Escobar A: Neonatal tetanus mortality in Veracruz 1989. PAHO Bull (accepted). 0 I I I I I I I I I I I /

I I I I I I I I I I I I I I ENVIRONMENTAL TOBACCO SMOKE AND LUNG CANCER MORTALITY IN TIlE AMERICAN CANCER SOCIATY'S CANCER PREVENTION STUDY II by Victor M. Cardenas, M.D., National Autonomous University of Mexico, 1978 Epidemiologist, School of Public Health of Mexico, 1981 M.P.H, Emory University, 1990 Adviser: Harland D. Austin, D. Sc., An Abstract of A dissertation submitted to the Faculty of the Graduat~ School of Emory University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Division of Epidemiology 1994 I

I In presenting this dissertation as a partial fulfillment of the requirements for an advanced degree from Emory University, I agree that the Library of the University shall made it available for inspection and circulation in accordance with its regulations governing materials of this type. I agree that permission to copy from, or to publish, this dissertation may be granted by the professor under whose direction it was written or, in his/her absence, by the Dean of the Graduate School when such copying or publication is solely for scholarly purposes and does not involve potential financial gain. It is understood that any copying from, or publication of, this dissertation which involves potential financial gain will not be allowed without written permission. Vfctor M. Cardenas 0 0~ 0.~ O~ 0 I I I I i I I I I I I I i I I

I I I I- I I I I I I I I I I I I I I I vii Abstract Background Evidence from epidemiologic studies and animal and genotoxicity assays leads to the conclusion that environmental tobacco smoke (ETS) causes lung cancer. Objectives The hypothesis of a causal relationship between exposure to ETS during adulthood and risk of death from lung cancer was tested using seven years of follow-up data of the American Cancer Society's Cancer Prevention Study II. Methods Three analytic cohorts of life-time nonsmokers were assembled for ETS analyses based upon: i) self-reported current hours of exposure to ETS, 2) exposure from spousal smoking, and 3) dose of exposure to ETS from cigarette smoking of spouses. These cohorts included 362, 265, and 127 lung cancer deaths, respectively. Mantel-Haenszel rate ratio analyses by ETS exposure variables, followed by Cox regression modeling controlled for age, gender, race, education, intake of vegetables, fruits and fat, occupational exposure to asbestos and history of chronic lung diseases. Results Lung cancer risk among nonsmokers was not associated with current self-reported number of hours of exposure to ETS. Nonsmokers reportedly exposed to ETS for 6+ hours a day had a 20% statistically not significant increased risk: the multivariate rate ratio (RR) was 1.2, [95% Confidence Interval (CI)=0.8-1.7]. Nonsmoking women married to current smokers had also a statistically not significant 30 percent greater risk of developing lung cancer (multivariate RR=I.3, 95% CI=0.8-l.9). Women married to current cigar/pipe smokers had a 50% increased risk, but again the finding was not statistically significant (multivariate RR= 1.5, 95% CI=0.8-2.7). After adjusting for all covariates, we found among nonsmoking women an increasing risk of lung cancer with increasing pack-years of cigarettes smoked by their husbands which was not statistically significant (p=0.14). Discussion Missing data on self-reported ETS could have resulted in considerable misclassification and thus biased the study findings towards an absent of an effect. Self-reported hours of ETS exposure does not measure intensity of exposure and referred only to current, rather than lifelong ETS exposure. Assessment of exposure to ETS using spousal smoking habits is better, but still has considerable limitations and at best is an imperfect measurement. 2"he study also has limited statistical power and its generally null findings are still consistent with the positive association reported from other

oo. i studies. Conclusions This study found no evidence of an association between self-reported ETS and lung cancer risk among nonsmokers. However, using spousal smoking habits to assess exposure, we found ETS is only weakly, and not statistically significantly, related to lung cancer risk among nonsmoking women in seven years of follow-up of the CPS II cohort. I I I I I i I f I I I I I I I I I

I I ! I. I I I I I I I I I I I I I I Chapter 1: Introduction 1.1. Background Epidemiologic studies, along w~th collateral evidence from experimental studies in animals and genotoxicity assays, indicate that tobacco smoking causes: I) lung cancer, 2) other cancers (oral, laryngeal, esophageal, bladder, renal, pancreatic, stomach, cervical and leukemia); 3) acute and other chronic conditions such as hypertension, coronary heart disease, peripheral arterial occlusive disease, cerebrovascular disease, chronic obstructive pulmonary disease, gastrointestinal disease, disease of the mouth; 4) that parental smoking is associated with major reproductive disorders such as low birthweight, birth defects, and the fetal tobacco s}ndrome; and 5) an increase in overall mortality (US DI-IHS Surgeon General 1989). Three decades after the publication of three seminal reports of case-control studies on the association of active smoking and lung cancer (Wynder 1950, Doll 1950, Levine 1950), the first two reports from epidemiologic studies on the effect of passive smoking and lung cancer among nonsmokers were published (Hirayama 1981,Trichopoulos 1981). In the 12 years since those reports were published, the scientific evidence accumulated on the effects of environmental tobacco smoke (ETS) on health, is not as strong as for active smoking. However, with respect to the potential health hazards from ETS, it seems that in recent years, a consensus is forming that passive smoking is injurious to health, although to a lesser degree than smoking. In particular, a report from the US National Research Council concluded after reviewing 13 published epidemiologic studies, that "a summary estimate from epidemiologic studies places the increased risk of lung cancer in nonsmokers married to smokers compared with nonsmokers married to non-smokers at about 34 percent" (NRC 1986). The International Agency for Research on Cancer concluded: "knowledge of the nature of sidestream and mainstream smoke, of the materials absorbed during 'passive smoking', and of quantitative relationships 0

I between dose and effect that are commonly observed from exposure to carcinogens leads to the conclusion that passive smoking gives rise to some risk of cancer" (O'Neill 1987). A recently published report from the US Environmental Protection Agency, included evidence from a total of 29 epidemiologic studies on the effect of ETS on lung cancer risk and concluded that "ETS is a Group A human carcinogen, the classification used only when there is sufficient evidence from epidemiologic studies to support a causal association between exposure to the agents and cancer" (EPA 1992). Since the late 1980's smoking in many public places, as well as on local and international airplane flights, has been banned. As the wave of anti-smoking campaigns rises, banning of smoking in public places and workplaces will further contribute to the goal of a tobacco-free world. Epidemiologic research on this subject is needed since there is still controversy over the size of the effect of exposure to ETS on the risk of cancer and other diseases, and whether the observed modest increased risk is causal or an artifact due to: 1) misclassification of smoking status, described by Mantel and Lee (Mantel 1983, Lee 1984); 2) a publication bias (Mantel 1990, Vandenbroucke 1988), or 3) confounding by socioeconomic status or other unspecified/unknown factors (Mantel 1992). The EPA report (1992) and a study and paper by- Fontham et al. (1991) offer reasonable rebuttal of these critiques. 1.2. Specific Aim The purpose of this study is to assess, in a large prospective study, whether ETS exposure in adulthood causes lung cancer. The objective of this study is to quantify the effect, if any, of adult ETS exposure on lung cancer mortality among lifelong nonsmokers in the American Cancer Society's Cancer Prevention Study II (CPS I13. 0 I I I I I I I | I I I ° i I

I I I I I I I I I I I I I I I I I I I 3 1.3. Hypotheses The specific research questions that were tested in this study can be stated, in the alternative form, as follows: l 1. Non-smokers exposed to ETS (either spousal or self-reported exposure) are at higher risk of fatal lung cancer than are non-smokers not exposed to ETS. I 2. The risk increases in a dose-response relationship with: i) hours (per day) of self-reported ETS exposure (cumulative at home, work or other places), and 2) pack-years of cigarettes smoked by spouses (married once and that had complete smoking data). 3. The relative risk of lung cancer among nonsmokers exposed both to ETS and occupationally exposed to asbestos, is greater than the risk of those exposed to I ETS only. ' 4. The association remains after adjustment for relevant potential confounders, and is not attributable entirely to misclassification of smoking status (i.e, misclassified smokers are included into a study restricted to nonsmokers). 1.4. Definitions of ETS, Lung Cancer, and ETS in General Analytical Cohorts The term ETS refers to "aged exhaled mainstream smoke (MS) from the smoker, diluted sidestream smoke (SS) emitted from the smoldering tobacco between puffs, contaminants emitted into the ah" during the puff, and contaminants that diffuse through the cigarette paper" (EPA 1992). In addition, it has been pointed out that ETS from pipe and cigar smoking should be considered in assessing the effect of ETS (Lychou 1986, Pershagen 1986). Few studies

4 (e.g., Garfinkel t985, Fontham 199 I), have assessed the risk of lung cancer I among nonsmokers by type of smoking of spouses. Although both smokers and nonsmokers are exposed to ETS, the effect of active smoking on lung cancer risk would overshadow any small effect of ETS among smokers, and hence most studies, as is this one, are restricted to the nonsmoking population. Passive smoking is a term that refers to inhalation of ETS by a non-smoker exposed to a smoking environment. The term "involuntary smoking" has the same connotation as passive smoking. These terms, along with "inhalation of second-hand smoke", are no longer in wide use in the recent literature, and we will only refer hereafter to ETS.. ETS Exposure in CPS "II The information on ETS in the CPS ri" cohort includes: A. Self reported number of hours exposed to ETS: the average number of hours subjects were reportedly exposed to "the smoke of others" at home, work, and other places, separately, for each of these settings and in total (all settings combined). B. Spousal ETS exposure: Smoking history as reported by the spouses of nonsmoking subjects, including the type of smoking habits (cigarettes, cigars, pipes), if current or former smokers, and the cumulative exposure to ETS from spousal cigarette-smoking during marriage. Lung Cancer in General Lung cancer is a group of malignant neoplasms that arise from: 1) the bronchial or bronchioloalveolar surface epithelium, 2) the bronchial mucous glands, or 3) a combination of the previous. According to the International Classification of (. I I I I I I I I I I I I I i I

I I I I I I I I I I I I I I Diseases Ninth revision (ICD-9~ WHO 1975), primary cancers of trachea, bronchus, and lung are grouped under the ICD-9 code N 162.0 to N 162.9. Lung Cancer in CPS II This is a study of mortality from lung cancer. Therefore, the outcomes refer to deaths for which the underlying cause of death selected by nosoloNst was coded N 162. I Self-reported and Spousal Exposure to ETS This terminology distinguishes t~e study population for which self-reported average number of hours of ETS exposure is used as exposure variable (self- reported ETS) from the population of nonsmoking wives and husbands in CPS II (spousal ETS) that uses spousal exposure data. Throughout the text other standard epidemiologic nomenclature is used (Last 1988). The glossary at the end of the text contains an exhaustive list of abbreviations used below. 1.5. Significance and Relevance of the Study CPS II provides a good opportunity to test the ETS-lung cancer hypothesis, with numerous advantages over existing epidemiologic evidence. First, the study provides information on 392,226 non-smokers, on 362 deaths from lung cancer among nonsmokers for analyses of self-reported ETS and 265 lung cancer deaths for analyses of spousal ETS, and 127 lung cancer deaths for dose- response analyses, numbers nearly as large as those of the largest case-control studies (Gao, 1987, Fontham 1991, Brownson 1992) or any other cohort study, including the CPS I (Garfinkel 1983), on this issue. Only four other cohort studies (Hirayama 1981, Hole 1989, Butler 1988) have addressed this issue; of I

6 I I I them only the CPS I study (Garfinkel 1983) had a statistical power greater than 80 percent of detecting a 1.5 increased rate. Second, the prospective nature of the study limits the possibility of recall bias, since information on exposure was obtained at the beginning of the follow-up. Prevalent cancer cases (i.e., persons with cancer at baseIine) were also excluded, limiting the possibility of recall bias. Th]rd, the information on both active smoking and ETS is unusually detailed and includes both: 1) self-reported average number of hours a person is exposed to the smoke of others (at home, work and elsewhere) and 2) smoking habits of spouses. Spousal smoking provides an important cross-check of the validity of self-reported smoking data. In most cases the spousal smoking information was provided directly by the respondents, unlike many case-control studies which rely on surrogate respondents. CPS IT shares this advantage with four smaller cohort studies (Garfinkel 1983, Hirayama 1981, Hole 1989, Butler 1988). Also, this study provided information to assess the effect of potential bias by active smoking status misclassification (Mantel 1983, Lee 1984). To set the limits of a reasonable bias by misclassification of active smoking status, we needed to obtain estimates of: l) the concordance of smoking status among spouses (i.e., smoking status as reported by spouses and by study subjects themselves); 2) the association of smoking status, especially of former smoking, a group over-represented among misclassified nonsmokers, and lung cancer; and 3) of the validity of the classification of smoking status. CPS IT provides information on smoking habits of members of the same household therefore, the concordance of smoking status among spouse-pairs can be estimated. This along with published estimates of misclassification of active smoking in CPS I and of the association of different cell-types of lung cancer with active smoking provided the necessary data to set limits to potential bias from misclassification of active smoking status. PO 0 O~ O~ 0 Ol I I I I I I I I <! I I I I I

I I I I- I I I I I I I I I I I I I I I I At the same time, this study assessed potential confounders of the relationship of interest such as age, diet, history of chronic lung disease, and exposure to occupational risk factors such as asbestos exposure, as well as concomitants of socioeconomic status such as education and 'race'. / Finally, this is an important scientific issue and is a subject of ongoing epiderrdologic research projects (Coleman 1992). Some ongoing projects on this topic are: I) a large intemation collaborative study being conducted in Western Europe and the US; .) a study planned to collect data from 400 lung cancer cases among nonsmokers in Russia; 3) in the US, two studies, one in Michigan and the other in New York, are in progress. 1 1.6. Extent of ETS Exposure Estimates of prevalence of exposure were not available until very recently. The 1988 National Health Interview Survey (NCHS 1988) included a set of I questions for adults about their lifetime working status and their work experience in the year before the interview. ETS exposure information was obtained from 44, 233 respondents. Interviewees were asked: "Do you live with a smoker?" and "Do they smoke at home?". According to the 1988 NHIS, about one / quarter of adults live with at lea~t one smoker at home, and almost 90 percent of smokers smoke at home. Some of the data obtained in this survey are presented in Table 1. Preliminary results from the ongoing National Health and Nutrition Examination Survey I-i-I (NHANES II1"), in which serum levels of cotin.ine are measured to assess exposure to tobacco smoke by persons in the US aged > 4 years, are available. The first 800 samples were tested with a highly specific test for serum cotinine, and all of them had measurable levels of cotinine. Results showed a bimodal distribution, with its antimode around 10-15 n~mL (MMWR 1993), the cutoff most often used to distin~ish smokers from nonsmokers. These data showed the ubiquitious nature of ETS exposure in the US. These preliminary results from NHANES III agree with a previous report by Wald on levels of

serum cotinine among 101 nonsmoking men, who had a mean of 8.5 nffmL (with a standard deviation of 1.3) (Watd 1984). A study of 380 nonsmoking participants in a cancer screening program in Buffalo, New York, conducted in 1986, showed that 24.3 percent of men were exposed to the ETS of their wives, whereas 66.0 of the women were married to smoking men. About 70 percent of the participants reported some exposure at home, and exposure at work was reported by 87 percent of subjects (Cummings 1989). Also, the 1988 NHIS included questions about policies at worksites and exposure to ETS at work, on which a report has been published (MMWR 1992). Half of the smokers in the 1988 NHIS reported some discomfort at work because of the smoke of others, whereas 84 percent of the nonsmokers reported some discomfort from ETS at their workplace. Thus, from. a broad public health perspective, this study has important implications. First, exposure to ETS is preventable by means of regulations at work sites and public places, as has been shown in the US (Fielding 1992). Smoking control measures are implemented by the government and the private sector in the US, and increasingly in the rest of the world. By 1987, in the US, restrictions were in place in more than 42 States and the District of Columbia, for smoking in transportation facilities, hospitals, schools, elevators, government buildings, and recreational facilities. According to the 1988 NHIS data, 40.3 percent of the 114.1 million employed adults in 1988 (who reported that their workplace was not their home), worked in locations where smoking was allowed only in designated areas. This estimate is consistent with results from the 1986 Adult Use of Tobacco Survey, which reported that 42 percent of worksites had restrictive policies (MMWR 1988). An estimate derived from the National Survey of Worksite Health Promotions, showed that 76 percent of work sites with smoking cessation activities had a smoking policy in effect (Fielding 1992). 0 O~ IX) 0 I I I I I I i I I

I I I I- I I I I I I I I I I I I I I Table 1. Percent distribution of persons 18 years of age and over by cohabitees smoking status, according to gender, . education, occupation, race, and age; United States, 1988. I Live with smoker Characteristics Yes % No % Unknown % Sample Size ..... Total Gender Male Female Years of education <12 13-15 16+ Occupation White collar Blue collar/service Blue collar/other Other Race Black Other Age 18-29 3O 4~. 45-64 26.6 71.5 1.8 44,233 26.2 71.9 1.9 18,562 I 27.0 71.2 1.8 25,671 30.6 67.6 1.8 25,671 28.3 70.0 1.7 9,808 17.8 80.3 1.9 10,990 122.8 75.3 1.9 22,505 33.1 65.2 1.6 6,535 30.1 68.3 1.6 13,169 26.2 70.2 3.6 2,024 26.5 71.8 1.7 36,864 I 28.3 69.3 2.4 6,186 25.2 72.6 2.2 1,183 33.6 64.6 1.8 10,5 I6 27.2 71.1 1.7 13,987 27.4 70.9 i .8 10,747 65+ _,, I 13,0 ~5.0 2.0 8,983 Source: cD~ N~HS: 0~'cupational Health SUpplement to the 1988, NHIS. ' The US Public Health Service Centers for Disease Controt and Prevention's National Institute for Occupational Safety and Health has recently released a ,, recommendation to reduce' ETS in the workplace to the lowest feasible concentration, and suggested to employers "to minimize the occupational exposure to ETS by all available preventive measures" ~ 1991). 1.7 Organization of Thesis In this chapter, we described the aims, specific objectives, and relevance of this study. Definitions of ETS exposure and lung cancer were provided. Chapter 2, deals with the epidemiology of lung cancer, and the role of smoking as a

l0 I I cause of cancer, as well as other risk factors. The biological plausibility of the carcinogenicity of ETS is reviewed. Special emphasis is placed on reviewing the literature on ETS and lung cancer; a summary estimate of 36 studies is presented and compared with that of the 1992 EPA report. These data are used to illustrate a funneling of estimates by sample size and to assess the potential of publication bias in studies of ETS and lung cancer. The lirnitations of existing studies on this issue are reviewed. Chapter 3 narrates the methods used in this study, with special reference to the selection of the analytic cohort, specification of exposure and outcome variables, and rationale and description of the statistical modeling used. Also a brief description of the CPS II design is given. The steps followed to validate death certificate data on lung cancer, as well as self-reported exposure to ETS are presented in a separate section of that Chapter. Chapter 4 gives demographic descriptive information for the CPS rr entire cohort, the analytic cohort based on self-reported ETS and the analytic cohort based upon spousal ETS and gives a comparison with the entire US population. Chapter 5 presents the results of the validation studies to use lung cancer as underlying cause of death from death certificates.as diagnosis of lung cancer, as well as the validity of self-reported exposure to ETS. Chapter 6 provides a description of the exposure variables in the entire CPS 12 population, and analytical cohorts. Chapter 7 presents the main results of the study based upon the self-reported exposure to ETS at home, work, and other places, and ETS spousal exposure. The relationship of ETS with potential confounders and effect modifiers is also presented. Chapter 8 includes the general discussion and conclusions of the study. The reader will find tables and figures incorporated into the text. Detailed tabular data of existing studies on ETS and lung cancer are given in Appendices, along with the CPS II questionnaires. r~ o o~ 0", o oi o~ I I I I I I I

I I I I- I I I I I I I I I I I I I I ll Chapter 2: 1 Literature Review and Metanalysis 2.1. Descriptive Lung Cancer Epidemiology I Lung cancers account for 15 percent (or 168,000) of all new cancer cases, and 28 percent of all cancer deaths (or 149,t300) in the US (ACS 1993). According to most recently published US mortality statistics, the lung cancer epidemic may have reached its peak among oldel: men (Boring 1993) and is declining among young men and young women (Devesa 1989, Glass 1991). The overall incidence rate based on the National Cancer Institute (NCI) Surveillance, Epidemiology and End Results (SEER) data was 57 per 100,000 in 1984-i988 (NCI SEER 1991). Whereas both ~cidence and mortality rates from lung cancer have begun to decrease in men, they continue to increase in women (all ages combined), and lung cancer has now surpassed breast cancer as the most comon cause of death from cancer in women. These changes in the epidemiology of lung cancer are reflected in the ~ to female ratio among new cases; it was 2: I in 1984-1988, but it was 6:1 in the 1960's (Schottenfeld 1975). Age-specific incidence rates of lung cancer increase exponentially with age. This observation has been considered a function of duration of smoking (Doll 1978), although it is argued that there is an independent effect of age at uptake of smoking habits (Molgaavkar 1989). Cohort analysis of lung cancer indicates that incidence peaked among men born between 1925-1930 and among women born between 1935-1940 and declined in subsequent birth cohorts, a pattern that mirrors the changes in the prevalence of cigarette smoking (Devesa 1984, 1989). In the SEER Cancer Registries data, lung cancer incidence in 1984-1988 was 37 percent greater among blacks (both males and females) than whites (SEER 1991). The race difference has widened since 1969 when the Third National Cancer Survey (Wynder 1975) reported an 11 percent excess for black men compared to white men and no excess for black women in relation to white women, a difference that reflects changes in smoking patterns. O' 0~ 0~ 0

12 I According to the same sources, an increase in the 5-year survival of lung cancer patients from 4 percent to 13 percent among whites and from 5 percent to 11 percent among blacks, was observed between 1950 and 1987 (SEER 1991). However, the overall survival of these cancers remains poor, with a median survival of less than a year from diagnosis (Wetzel 1989). Few reports exist on specific survival rates by stage at diagnosis; in one of them, comprising 999 registered cases at the Northwestern Memorial Hospital from 1976 to 1985, 48.7 percent of the cases had distant spread at the time of diagnosis. Even among patients with localized disease, median survival was only 22 months (Wetzel 1989). This poor survival, in conjunction with secular changes in smoking prevalence and better breast cancer survival rates, explains why, among US women, lung cancer has surpassed breast cancer as the leading cause of cancer mortality among women (Boring 1992). 2.2 Smoking and Lung Cancer The earliest observations of the association of tobacco and cancer can be traced to 1761 in a report by John Hill, who described cancer of the nose among users of tobacco snuff (Redmond 1970). A literature review by Adler, reported in 1912, could document only 374 lung cancer cases worldwide, and the author suspected that cancer of the lung was decidedly on the increase. In 1918, Yamagiwa and Ichikawa reported on the experimental induction of squamous cell skin cancer by application of coal tar in the rabbit model (YamaNwa 1918), confirming the early observations of Percival Pott upon scrotal cancer among chimney-sweepers (Shimkin 1975), and thus lending plausibility to the cigarette smoking-lung cancer hypothesis. Specifically the association between tobacco smoking and lung cancer was f'u-'st reported in Germany based upon clinical and autopsy evidence by Lickint and Hanf in 1928, (Vincent, 1986) and further confirmed by Mtiller in 1939 (Wynder, 1975). However, it was not until the early 1950's that the major link was established through epidemiologic studies. Following the publication in 1950 of reports from case-control studies describing a strong positive association of smoking and lung cancer (Wynder 1950, Doll 1950, Levine 1950), numerous other epidemiologic studies have consistently 0 0 0 I I I I ! I I ,I I <I I <I I I

! ! I I- I I I I I I I I I I I I I I 13 reported increased incidence land mortality from lung cancer among smokers compared with non-smokers, with a "best" estimate of the relative risk of 22 in men and 12 in women (US Surgeon General 1989). Tobacco smoking also is associated with cancer of five other sites (larynx, buccal cavity, pharynx, ! esophagus, and bladder), and tl~ough not fimaly established, it may be a cause of cancers of the pancreas, kidney, stomach, cervix and leukemia (Thomas 1992, Garfmkel 1990). More than 50 independent case-control studies and eight cohort studies consistently reported associations between smoking habits and overall, .! lung cancer, other upper respiratory and digestive cancers, chronic bronchitis, coronary disease, and peptic ulcer mortality. Smoking is also responsible for an excess of deaths from cardiovascular diseases that until recendy, surpassed the magnitude of smoking-attributable lung cancer mortality in the US (Shopland • 1991). Smoking also ~ncreases blood pressure. Also, parental smoking is associated with major reproductive disorders such as low birthweight, including the so-called fetal tobacco syndrome (Thomas 1992), and birth defects (Kelsey 1978). Last, smoking is a cause of several gastrointestinal and mouth diseases, such as peptic ulcer disease and' ginNvitis (Thomas 1992). 2.3. Biological Plausibility More than 4000 chemical compounds have been identified in ETS (Surgeon General 1986, IARC 1987, NRC 1986, EPA 1992, O'Neill 1987, L~Sfroth 1989, Claxton 1989), of which eighteen are known carcinogens: benzene, formaldehyde, hydrazine, N-nitrosodimethylamine, N-nitrosopyrrolidine, 2- toluidine, 2-naphtylamine, 4-an~obiphenyl, benz[c~]anthracene, benz[o:lpyrene, "~'-butyrolactone, quinoline, N'-nitrosonornicotine, NNK [or 4-(N-methyl-N- nitrosoamino)- 1-(3-pyridyl)- 1-butanone], N-nitrosodiethanolamine, cadmium, nickel, and 210 Polonium. Five studies measuring personal exposure to particulate matter associated ~vith ETS for nonsmokers, and another five on exposure to airborne nicotine associated with ETS, were reviewed in a recently published report of the Environmental Protection Agency (EPA 1992). Accordingly, particle mass due to ETS in the respiration zone of nonsmokers ranges from 18.4 to 64 gg/m3, and 0.1 to 40 gg/m3 of nicotine.

14 I I I The size of the SS particles is smaller and dilutes more rapidly in air, than MS particles. Some carcinogens, like nitrosamines, benzo[a]pyrene, benzene, cadmium, nickel and aromatic amines, are found in larger quantities in SS than in MS. The NRC reports pointed out that "constituents of the vapor phase such as N-nitrosamines would be more likely to remain in the ambient air for longer spans of time ". For instance, two constituents of the vapor phase are N- nitrosodimethylamine and benza[a]pyrene. They are found at a 20 to I00 SS /MS ratio (range in SS: 200-4000 ng), and a 2-4 ratio (range in SS: 40-280 ng), respectively. Smoke of cigars and pipes carries more carcinogens than do cigarettes, in addition to the obvious observation that they produce smoke in larger volumes than do cigarettes. In particular, the smoke of cigars contains more benz[a]pyrene and pyridine, whereas that of pipes has more tar (Shephard1982, Appel 1990). A model-based approach (Repace 1980, Repace 1982) was used in a report of the National Research Council (NRC 1986) to describe ETS exposures. As illustration of this model, the NRC report presented a range of 10 to 100 total respirable particulate matter emitted (l.tg/m3) "in a residence with one smoker smoking at a rate of either i or 2 cigarettes per hour for the range of mixing, ventilation, and removal rates occurring in residences under steady-state conditions" (NRC, 1986). 2.4 Epidemiological Evidence Two seminal reports (Hi.rayama, 1981, Trichopoulos 1981) implicated ETS as a. risk factor for lung cancer among non-smokers. These two, along with 34 other reports are summarized in Table 2. A detailed review of most of these studies can be found in the Appendix of the I992 EPA report, but it is worth summarizing their key features: I) without exception they are restricted to nonsmokers, 2) most of them have concentrated on women, 3) 32 are case- I I I I I I I ,! I I i I I

! I I I- I I I I I I I I I I I I I I 15 I control studies, 4) ETS exposure has been measured in terms of spousal smoking history reported by cases or next-of-kin, most of them have relied on exposure information from proxy respondents, and 5) most of them had little statistical power for detecting ~ weak association. The original reports from 36 studies were reviewed and we abstracted information on the association between having a smoking spouse and the occurrence of lung cancer. ~onfidence limits for the studies in Table 2 are shown in Figure 1, arranged by sample size. Nine of these 36 studies reported a positive, statistically si~cant association; in three other instances a positive and borderline significant association was found (i.e., point estimate above one and lower 95% confidence limit----0.9). A negative point estimate was reported in eight studies, but none of these negative studies was statistically significant. Only twelve studies had at least a 50 percent statistical power (i.e., information on at least i00 lung cancer cases or deaths among nonsmokers) of detecting a ris~ ratio of 1.34, the hypothesized size of the ETS effect on lung cancer risk in the NRC meta-analysis. A summary estimate over these 36 studies was obtained using precision-based estimators (i.e., weighting eac study by the inverse of the variance of the RR on the logarithmic scale): ETS during adulthood (i.e., married to smokino spouse) increases the risk of lung cancer among nonsmokers by 20 percent (95 percent confidence interval (95% CI): 1.1-1.3) (See Appendix A, for 2x2 tables of these studies and calculations). Mor.eover, in most studies, even in those who did not I find an overall association, there was evidence of a dose-response relationship between ETS exposure and the risk of lung cancer among nonsmokers. The funnel plot in figure 2, shows the inverse relationship between precision and sample size. Vandenbroucke land Mantel have argued that there is some evidence of a publication bias, since there is a deficit of small negative studies, depicted in figure 2 in the lower left of the funnel plot (i.e., the non-significant small negative studies) (Mantel t990, Vandenbroucke t988). However, four small negative studies are shown inI this plot. Even if this deficit exists it would be

16 largely overwhelmed by the number of both positive and negative large studies in the right section of the plot. A closer examination of studies by country, as reported in the 1992 EPA report, shows that two large studies in China (Wu- Williams 1990, and Gao 1987) reported statistically significant inverse associations, whereas those in Japan and Europe were more likely to be positive and to report stronger associations as compared to US studies. Our metanalysis summary estimate of the RR is remarkably close to that of the EPA report on US studies, which is 1.19. The EPA report made a downward adjustment by misclassification bias and reduced the observed associations. Since on average that adjustment represented less than 10 percent of the point estimates, ours and those of the EPA report are approximately the same. 2.5 Collateral Evidene~ In support of the role of ETS as cause of lung cancer among nonsmokers, a recently published autopsy-based study (Trichopoulos 1992) documented an increase of pre-cancerous lesions due to ETS, opening a new research avenue on this issue. Another piece of epidemiological evidence that supports the claim that ETS causes lung cancer derives from a study of lung cancer among dogs in relation to the smoking habits of their owners. The authors found an association of magnitude similar to that reported in humans for ETS and lung cancer risk. Interestingly, also the authors noted that the association held for brachicephalic/mesocephalic dogs but not for dolicocephalic dogs, a fact interpreted by the researchers consistent with an effect from exposure to volatile smoke particles as those of ETS (Reif 1992). This study reinforces the findings of experimental studies of lung tumors among male beagles trained to smoke through a tracheostomy: the authors concluded that smoking of cigarettes greatly increased the development of such tumors (Auerbach 1970). I I I I I I I l I I I I I i I I

I I I I- I I I I I I I I I I I I I I 17 I Table 2 Summary of 36 epidemiologic studies on ETS and lung cancer risk, from 1981 to 1993. Point estimates and 95% confidence intervals of the association of |un,~ can~er and spouse'smoking habits No Author Design Study Weight Estimated Source Sizel 2 RR 1. Hirayama Cohort t74deaths L42.0 L.5 (t.L-2.l) Lancet L981 2. Garfinkel Cohort 153 deaths 88.0 1.2 (0.9-1.4) JNCI 1983 3. Hole et al. Cohort [ 9 deaths 7.0 2. I(0.5-14.7) BMJ 1989 4. Butler Cohort 8 deaths 2.0 2.0(0.6-6.7) UCLA 1988 5. Trichopoulosetal Case-control 77cases 12.7 2.1(L.2-3.7) Lancet1983 6. Chart & Fung Case-control 86 cases 12.8 0.8(0.4-1.3) Grundmann, 1982 7. Correa et al. Case-control 32 cases 6.8 3.1(1.5-6.8) Lancet 1983 8. Buffler et al. Case-control 52 cases 8.5 0.8(0.4-1.6) Mizell 1983 9. Dahlager et al. Case-control 48 eases 8.3 1.5(0.8-2.8) Cancer Res 1986 I0. Kabat & Wynder Case-control 36 eases 4.5 0.9(0.3-2.2) Cancer 1984 I I. Garfinkel et al. Case-control 134 cases 22.2 1.2(0.8-1.9) J'NCI 1985 12. Wu et al. Case-control 31 cases 7.1 1.2(0.6-2.5) JNCI' 1985 13. Akiba et al. Case-control 113 cases 19.7 1.9(1.2-3.0) Cancer Res 1986 14. Lee et al. Case-control 47 cases 7.3 1.l(0.5-2.3) B.IC 1986 15. Gao et al. Case-control 436 eases 61.2 0.8(0.6-I.0) LIC 1987 16. Koo et al. Case-cont~[ol 86 cases 12.9 1.5(0.9-2.7) LIC 1987 17. Pershagen et al. Case-control 77 cases 16.4 1.2(0.7-2.1) AlE 1987 18. Humble et al. Case-control 28 cases 5.3 3.2(1.4-7.9) AJ-PH 1987 19. Browson et al. Case-control 19 cases 2.0 1.8(0.4-7.5) AYE 1987 20. Lam et al. Case-control 199 cases 30.6 1.6(1.2-2.3) BJ'C 1987 21 Lain & Cheng Case-control 60cases I0.I 2.0(I.1-3.8) Lee, 1992 22. Shimizu et al. Case-control 90cases 14.1 1.1(0.7-1.9) TJEM 1988 23. Inoue Case-cont](oi 22 cases 2.9 2.6(0.8-9.9) Smk& Hlth 1987 24. Geng et al. Case-control 54cases 8.1 2.2(1.1-4.3) Sink& Hlth 1987 25. Svensson et al. Case-control 34 cases 6.0 1.3(0.6-2.9) Acta Oncol 1989 26. Ianerich et al. Case-control 191 cases 14.0 0.9(0.6-1.6) NEJM 1990 27. Stockwell et al. Case-control 210 cases 9.6 1.6(0.8-3.0) JNCI,1992 28. Kalandidietal. C, ase-contml 91 cases ll.2 1.6(0.9-2.8) Can CaCti 1991 29. Sobue et al Case-control 144 cases 29.6 1.1(0.8-1.6) Ga No Rin 1990 30. Katada et al. Case-cont/ol 17 cases 0.4 NC(0.6- NC) Ga No Rin 1988 31. Wu -Williams Case-control 417 cases 61.3 0.8(0.6-1.0) B5C,1990 32. Kabat et al. Case-control 89 cases 16.0 1.0(0.6-1.7) Lee 1992 33. Liu et al. Case-control 54 cases 5.6 0.7(0.3-1.8) ISE 1991 34. Fontham et al. Case-control 420 cases 59.4 1.4(1.1-1.8) CancerEpid 1991 35. Browson et al. Case-control 451 cases 78.6 1.0(0.8-l.2) AYPH 1992 36. Liu et al. Case-control 38 cases 5.7 1.7(0.7-3.8) AYE 1993 Total 4,227 802.6 1.2( 1. l- 1.3) tNumber of lung cancer cases; 2, Inverse of the variance of the logarithm of the RR estimate o~ o-

I -. Mct~nelysts 422"7 GJo CC ~ F~m CC 4~ Wu CC ~ ~C l~ J~ch CCl~ Hi~ytm8 C~ 174 ~e CC I~ Ga~nkel CC !~ A~bt CC ~It~Mf CC 91 ~f~ CC ~ Ktbtt CC ~ ~a.f CC ~ K~ CC ~ T~ CC ~ ~e~t~. CC ~ ~ CC~ G~E CC 54 Liu CC ~ Dahla~ ~ ~CC~ Ksbs{ CC ~ Sv~ CC ~ C~a CC ~2 Wu CC ~I Humb~ CC ~ l~e CC ~ B~ CC 19 Kstads 17 Buder C~ 8 | I I I I I I I I I I I

I I " Figure 2. Funnel plot of relative risk (on log scale) according to the size of SD (In RR) of 36 studies of lung cancer and ETS I tspousal exposure 0.8 0.6 0.4 0.2 0 rl.5 • " " • ,.. ",",", "- . • l-0.5 .... , ---" "-'-|,-'...-;- --/ (RR) I ' " " = " I"0"5 In =-15 I I I I I I I I I I 2.5 Collateral Evidence SD [in (RR)] In support of the role of ETS as cause of lung cancer among nonsmokers, a recently published autopsy-based study (Trichopoulos .1992) documented an increase of pre-cancerous lesions due to ETS, opening a new research avenue on this issue. Another piece of epidemiological evidence that supports the claim that ETS causes lung cancer derives from a study of lung cancer among dogs in 1 relation to the smoking habits of their owners. The authors found an association of magnitude similar to that reported in humans for ETS and lung cancer risk. Interestingly, also the authors noted that the association held for brachicephalic/mesocephalic dogs but not for dolicocephalic dogs, a fact ! interpreted by the researchers consistent with an effect from exposure to volatile smoke particles as those of ETS (Reif 1992). This study reinforces the f'mdings of experimental studies of lung tumors among male beagles trained to smoke through a tracheostomy: the authors concluded that smoking of cigarettes greatly increased the development of such tumors (Auerbach 1970).

2.6 Controversy 2O More than 50 epidemiologic studies on the health consequences of exposure to ETS have been reviewed (NRC 1986, US Surgeon General 1986, Blot and Fraumeni 1986, IARC 1987, Letzel 1988, Wells 1988, Working Group on Passive Smoking 1990, Fleiss 1991, Lee 1992b, EPA 1992). Confounding, misclassification bias introduced by misclassification of smoking status, and a publication bias have been invoked to argue that the weak association of ETS with lung cancer risk is artifactual. Mantel and Lee have argued since 1984 (Mantel 1983, Mantel 1990, Mantel 1992, Lee 1984, 1986, 1988, 1992a 1992b, ) that the 20-50 percent excess of lung cancers among non-smokers married to smokers might be due to misclassification coupled with concordance of smoking habits between spouses. To illustrate how this misclassification might bias the study result, we have adapted from the NRC report, a hypothetical example shown in Figure 3. In this hypothetical situation, one million subjects are followed for 6 years and the occurrence of lung cancer deaths is ascertained. For simplicity, we assume that no entries or withdrawals other than deaths from lung cancer are allowed before the study ends. The cohort is analyzed as a closed cohort, and the parameter of interest is the risk ratio. Thirty-eight percent of this hypothetical population never smoked regularly, and the remaining comprise ever smokers. Misclassification of 3 percent of this population regarding thek smoking status results in an observed distribution of 60 percent of ever smokers and 40 percent of never smokers (Panel A of Figure 3). This percent of misclassified smokers is in agreement with published figures from CPS I fLee 1988). A study is then conducted among nonsmokers to determine if those exposed to spousal ETS increases the risk of lung cancer. Twenty-five thousand ever smokers get into this study of nonsmokers, and this is a key factor for the bias described here to take place (Panel B, Figure 3). The second key factor derives from the fact that smokers are more likely to marry smokers than to marry I I I I I I I I !1 I I

21 I I I I I I I I I I I I I I nonsmokers: in CPS II sixty-four percent of the smokers were married to smokers, and 46 percent of nonsmokers were married to nonsmokers. Another key factor for this misclassification bias to take place is a strong relationship of the outcome under study with active smoking: in this example, we assume a five- fold increased risk of death from lung cancer among smokers due to the fact that 70% of those deaths would be adenocarcinomas, hence an estimate of the relative risk of 4 for adenocarcinomas was used (Cfr. Brownson 1986), and weighted for 50% as former smokers, since most misclassified smokers would be actually former smokers (Cfr. Lee 1988). A reasonable estimate of the effect of smoking among misclassified smokers would be 5, because most of them would be former smokers. The rates of lung cancer death among nonsmokers in CPS II is 11 deaths per I00,000 (Garfinkel 1991). Knowing that only 3 percent of the population in this study is formed by smokers., using the relationship of total incidence, in this case mortality, to calculate the rates among the unexposed* : M0 = M, (Hermekens and Buring 1987) we arrived at the rates among truly classified nonsmokers. Therefore, the observed 7 percent increase is only due to misclassificafion of smoking status. The argument of bias from misclassification of active smoking status assumes that smoking spouses would have the same survival as nonsmokers. Indeed, subjects with a history of regular cigarette smoking have 24% (95% CI= 1.20- 1.28) increased risk of deaths from all causes as compared to never smokers (US DHHS: Surgeon General 1989). This assumption would work in the opposite direction of the misclassificafion bias. For instance, history of current ETS exposure among nonsmokers who were exposed in the past, might bias study results towards the null because of poorer survival of heavy smokers. * Notation M0=mortality among the unexposed; Ml=mortality among the exposed; Pc=proportion of the population exposed and P0=proporfion of the population unexposed. 0 0 I

22 i Figure 3. Hypothetical cohort study of" nonsmokers to show the effect of misclassification of smoking status on lung cancer rates by spousal ETS (adapted from NCR 1986). True distribution Observed distribudon under 3% misclassification 1,000,000 population 62% ever smokers and 38% never smokers Gold standard + Total Quesdormair~ . 25,000 375,000 400,000 Total 620,000 380,000 1,000,000 60 % ever smokers and 40 % never smokers Bo 1,000,000 population smokers: 620,000 smokers misclassificd as nonsmokers: 25,000 CPS II estimates of concordance 64% of smokers are married to smokers of smoking: 16,000 9,000 smokers nonsmokers 375,000 non smokers 46% of nonsmokers are married to nonsmokers 202,500 172,500 smokers nonsmokers RR of 5 cancer of the lung for active smoking assumming most misclassified smokers are truly former smokers Rate of lung 5.0 5.0 1.0 cancer 0.0004 0.0004 0.00009 death: (in six years) 38.4 21.6 109.4 1.0 0.00009 93.2 Do Observed study Cases xNN, x~ Spouse smoking status / Smoker Nonsmoker/ Person-Years 1,310,926 1,088,943 IX) o CO, 0", o o1 o~ o I I I I f I I I I I I I i

! ! I I- I I I I I I I I I I I I I I 23 In addition to the argument of bias from misclassification of acdve smoking, there are two other validity issues that have been brought into the discussion of the ETS=lung cancer hypothesis. Figure 2, presented a funnel plot that shows little indication of a publicad~)n bias. A similar plot based upont the published studies by 1988 led to argue that there was a publication bias on this issue (Vandenbroucke 1988). However, based upon the available information by 1994, it seems unlikely that those unpublished studies would have outnumberedbthe large positJve studies, and even less likely, that they would have outweighed the summary estimates such as those presented in reports by NRC, Fleiss, EPA and before in 2.4. Finally, some researchers hav(e suggested that the potential of confounding by some unspecified potential confounder such as socioeconomic status has been overlooked. For instance, it is argued that low socioeconomic status is associated with increased risk of cancer, and at the same dine, smoking is more prevalent among the poor. This.led Mantel to formulate the following hypothesis to explain the ETS-lung cancer association: (nonsmoking) "wives of smoking husbands would be affected by the concomitants of socioeconomic levels" (Mantel 1992). Most studies, however, have controlled for socioeconomic status or proxy variables of socioeconomic level, such as 'race' and education, and have still found an association between lung cancer and ETS. A recent review by Lee (1992b), concluded that: "Taken as a whole, the evidence reviewed does not demonstrate that exposure to environmental tobacco smoke increases risk of cancer, heart disease or other diseases among adult non- smokers." Four of the other major six reviews, however, agree that ETS is causally related to lung cancer, and we excerpted the following conclusions. The NRC report (NCR 1986) stated the misclassification bias pointed out by Lee "is not likely to account for all the increased risk." The Surgeon General's report (US DHHS Surgeon General I986) concluded "The absence of a threshold for respiratory carcinogenesis in active smoking, the presence of the 0 O~ O~ 0

24 I I I I I I same carcinogens in mainstream and sidestream smoke, the demonstrated uptake of tobacco smoke constituents by involuntary smokers, and the demonstration of an increased lung cancer risk in some populations with exposure to ETS lead to the conclusion that involuntary smoking is a cause of lung cancer." Blot and Fraumeni published a general review of the available epidemiolo~c evidence about the time of the publication of the NRC report, and provided a relative risk summary estimate of 1.3 (95% CI=I.I-1.5) (Blot and Fraumenl t986). Fleiss and Gross found the conclusions in the NRC report "unwarranted given the poor quality of the studies on which it (was) based." However, their assessment of nine US studies in the NRC meta-analysis, found no evidence of study-to-study heterogeneity, and is consistent with a small, although not statistically significant, increased risk (Fleiss 199 I). The effect of ETS on other diseases and adverse outcomes, is beyond the scope of this document, but it has been reviewed in detail in Working Group on Passive Smoking (Spitzer 1990). Risk assessment of ETS is a subject of many letters, editorials, articles and monographs in the scientific and medical literature (Mantel 1992, Steenland 1992, Glantz 1991, Vandenbroucke I988, Wells 1988, Glantz 1992, Lee 1992a, Heath 1993). 2.7 Other Risk Factors Ionizing radiation, including both a-rays, emitted by radon dust particles, and x- rays, asbestos, arsenic and nickel compounds, polycyclic aromatic hydrocarbons, hexavalent chromium, mustard gas, and other environmental and occupational exposures have been documented as risk factors for lung cancer (Blot 1984). A study by Selikoff, Hammond, and Churg showed that smoking and asbestos exposure have more than additive joint effects on the risk of lung cancer (Selikoff 1968). Previous lung diseases such as pneumonia, chronic bronchitis, asthma, and tuberculosis are known risk factors for lung cancer among nonsmokers (Alavanja 1992). Studies conducted in the 1980's using 0 0", I',0. o I I I I I I

I I I I- I I I I I I I I I I I I I I 25 i sera banks from prospective studies explored the effect of different nutrients on the subsequent risk of cancer (i.e., nested case-control studies). Lower levels of serum 13-carotene were found among persons who subsequently became cases than among controls as recentl~ reviewed by Comstock et al. (Comstock I992). Fontham reviewed dietary studies on this issue and reported that they have been notably consistent in finding an approximate 50% reduction in risk associated with high, compared with low consumption of carotene containing fruits and vegetables ('Fontham 1990). ~3ietary fat intake has been implicated as a risk factor for lung cancer in one large ease-control study (Alavanja, 1993). Empirical evidence coming from case-control studies reporting an apparent effect of family history of cancer; pedigree studies, variations in carcinogen- metabolizing enzymes and chromosomal markers are also consistent with the hypothesis of inherited susceptibility (Mulvihill 1984, Kellerman et al. 1973, Weston et al. 199I, Caporaso et al. 1990). Three studies of twins have shown a familial proclivity to smoking (Mulvihill 1984). I 2.8 Lung Cancer Classification Tumors of the respiratory tract include neoplasms of the oropharynx, larynx, trachea, lungs, and pleura. Epithelial malignant tumors comprise, according to the World Health Organization, the following major histological types of lung tumors: 1) squamous cell carcinoma, formerly called epidermoid carcinomas, 2) adenocarcinomas, 3) small cell carcinoma, including oat cell carcinoma, and 4) large cell carcinoma (Sobin, !1981). Estimates of the proportion each type rep.resents vary depending on the source data (i.e., biopsy cytology, surgical specimens, autopsy), and range from 33-64 percent, 16-26 percent, 9-20 percent and 19-25 percent respectively, of all malignant pulmonary neoplasms (Minna 1989, NCI SEER 1991). In~1962, K.reyberg divided lung cancer into two groups according to the strength of its association with active smoking (Kreyberg 1962). This classification was based upon observation recorded in the cohort study of British physicians (Doll 1957). Squamous cell, small cell carcinomas, and large cell carcinomas fel~ into the Kreyberg type I for those who had a

26 ! ! I I I stronger association with smoking. Kreyberg type II included adenocarcinomas and mucous-gland tumors, frequently diagnosed among non-smokers and women. A so-called third Kreyberg ~oup, showing intermediate strength of association with active smoking, includes adenosquamous, anaplasdc, and undifferentiated carcinomas. A recendy published large muldcenter case-control study of lung cancer in nofl-smoking women included histologically confirmation, and reported a more specific association of ETS with adenocarcinomas (78 percent of all cases) than with the other histologic types (Fontham 1991). Some previous studies, such as those by Kabat and Wynder, and CPS I (Garfinkel 1981), had found a smaller proportion of adenocarcinomas (i.e., 55% and 56%, respectively) (Kabat 1984). However, the vatidity of.the Kreyberg dichotomy has been increasingly questioned given the recent trends in lung cancer histopathology. 2.9 Measurement of ETS Exposure Ideally ETS should be measured directly using 1) air samples of ETS pollutants in various settings (home, work, and other places) or 2) highly specific biomarkers such as cotinine in saliva, blood and urine. Measurements of cotinine (a nicotine metabolite) in serum, urine and saliva have been used effectively to quantify exposure to tobacco smoke (Wall 1988). A highly speci.fie technique for serum cotinine has been developed recently to measure levels as low as 0.03 ng/mL (MMWR 1993). Today, questionnaires are the most commonly used method, and have unique advantages over direct measurements. The most important advatage is that questionnaires can describe past exposure that is relevant for diseases of long latency such as cancer. However, the extent of misclassification of self-reported ETS exposure may be extensive (Pron 1988). Questionnaires are also an inexpensive method that can be used in large' studies. Studies conducted to assess sources of ETS have consistently reported that ETS was ubiquitous at settings such as the workplace (Cummings 1989) and therefore 0 CO 0 0"~ i I I I i I

I I I i- I I I I I I I I I I I I I I 27 I exposure sustained in the workplace makes an important contribution to lifelong exposure. Two studies have reported that spouses are a very important source of ETS exposure and that by itself explains most of the total exposure (Becher 1992, Emmons 1992). Moreover, the validity and reproducibility of spousal ! ETS exposure assessment is greater than that of self reported ETS (Pmn 1988, Gann 1988). Another large study among ex-smokers and nonsmokers who were under the Kaiser-Permanente Medical Care Program was conducted to estimate and identify the frequency and determinants of ETS exposure: age was found I inversely related to ETS exposure, as was education to duration of self-reported ETS exposure (Friedman 1983). Evidence of denying of ETS exposure by subjects with less schooling was also found, leading the authors to suggest that "further effort be devoted to improving methods for assessing passive smoking ,, [ by questionnaire.. Nonsmokers with history of atopy or any respiratory illness were found more likely to report ETS exposure than subjects with no such history (Cummings 1991), implying that these subjects "are more likely to experience adverse acute reactions to ETS than people without such a medical history", suggesting individual differences in sensitivity to ETS. A study of the correlation of urine cotinine of infants with the smoking status of household members showed that this marker of tobacco exposure had a median / of 1.6 p.g/L for infants unexposed at home, that it was lower among infants living with smoking cohabitees but whose mothers were nonsmokers (median 8.9g.g/L) as compared to the levels of those infants whose mothers were the only smokers in the households (median 28 [.tg/L). In turn, the level of urinary cotinine among infants of smokin~ mothers who also had other smokers among the household members was even higher: 43 g.g/L (Chilmonczyk 1990). This study result underscores the importance of: 1) the number of smokers among cohabitees as source of exposure to ETS, and 2) of the relationship between family members who smoke and nonsmokers in determining the intensity of exposure to ETS in households.

28 I Chapter 3: Methods 3.1 The Cancer Prevention Study II Study Population: The Cancer Prevention Study II a) Recruitment CPS II is a cohort study of 1,185,I24 men and women whose mortality experience has been ascertained since 1982. CPS II is the third large prospective study sponsored by ACS. A key feature of the ACS studies is that volunteers invite family groups among their relatives, friends, and neighbors to participate in the study. The volunteers then assist in follow-up. The first study, often referred to as the Hammond-Horn study (Hammond 1958), was comprised of 188,000 white men 59-69 years-old living in 394 counties in nine states, recruited by 22,000 volunteers, and followed for 44 months, from 1952 to 1955. That study was a landmark in epidemiologic studies of cancer and provided compelling evidence for the causal role of active smoking on lung cancer and other diseases. In the next study, the Cancer Prevention Study I (CPS I), 68,I 16 ACS volunteers recruited a cohort of 1,078,894 men and women, aged 35-84 at enrollment and followed them over a 12-year period (Hammond 1966). Enrollment in CPS II began in September 1982 and was essentially completed by the end of November 1982. Approximately 77,000 ACS volunteers enrolled consenting families if at least one household member was 45 years or older, enrolled all family members who were 30 years or older. Enrollment of subjects was carried out in all 50 states, the District of Columbia, and Puerto Rico. Volunteers were asked to include families they thought would remain in the local area for the next six years (Stellman 1986). i I I I I '1

I I I I- I I I I I I I | I I I I I I 29 b) Follow-up The participants' vital status was determined using two approaches from the month of enrollment in theI Fall of 1982 through December 31, 1989. All volunteers made personal inquiries in September of 1984, 1986, and 1988 to determine whether their enrollees were alive or deceased and to record the date and place of all deaths. Since 1988 a new approach was used: automated linkage through the National Death ~dex (NDI) to extend follow-up through December 1989 (Calle 1993) and to identify deaths among 21,704 (1.8%) persons lost to follow-up between 1982 and 1988. By December 1989, 101,541 participants (8.6%) had died, 1,080,689 (,91.2%) were considered alive, and 2,894 (0.2%) I. had follow-up truncated on September 30, 1988. Specifically this group comprises persons who were followed by ACS volunteers through that point in time but who had insufficient data on names and date of birth to be sent to NDI for matching using the linkage system. Death certificates were obtained for 96.8 percent of persons known to! have died. Using the system described in the International Classification of Diseases, 9th Revision (ICD-9), a nosologist coded lung cancer deaths according to the ICD-9 code (WHO 1979). c) Baseline QuestionnaireI Persons enrolled in the study completed and returned a self-adrninistered, four- page confidential questionnaire that covered 400 items. Appendix C includes a copy of the CPS II questlonnatres for men and women. Baseline questions included personal identifiers, height, weight, demographic characteristics, personal and family history of cancer and other diseases; use of medicines and vitamins; occupational exposures; menstrual and reproductive history; diet and drinking habits; and other halbits, including active and passive smoking (See below). To classify the active smoking status, participants were asked the standard question: "Do you now or have you ever smoked cigarettes at least one a day for one year's time?". The questionnaire for men also inquired about cigar and pipe smoking. For every ~ype of active smoking, infomaation on number of

30 I I I cigarettes, cigars or pipes per day and duration of smoking habits was elicited for both current and former smokers. 3.2. Published Results from CPS II CPS II has akeady provided important information published in 32 papers in journals and book chapters, on different issues such as: 1) smoking trends and projected mortality from lung cancer in the US (US Surgeon General 198_9, Stellman 1988, Garfinkel 1991) and in economically developed countries (Peto 1992); 2) a protective effect of regular aspirin use on the risk of fatal colon cancer (Thun 1991), as well as the effect of dietary fiber (from vegetables and grains), physical activity, obesity and dietary fat on the risk of fatal colon cancer (Thun 1992); 3) the assessment of the risk of exposure to diesel exhaust (Boffetta 1988); 4) artificial sweetener use (Stellman 1988); 5) estrogen-related cancers and smoking (Garfinkel 1990); 6) leukemia and smoking (Garfinkel 1990); 7) the validation of follow-up procedures in CPS II through the National Death Index (Calle 1993), and 8) the relationship between hair dye use and fatal cancers (Thun 1994). 3.3 Main Design Features This is an ongoing prospective cohort study. By design, no new enrollees were allowed after 1982; therefore, CPS II is a closed cohort study. Individuals leave the cohort either because they die or because they are lost to follow-up. The mortality rate ratio is the parameter of interest (i.e., the measure of association of choice in this study), given the absence of incidence data in this study. In a cohort study, individuals contribute varying amounts of time under observation. Therefore, the statistical analysis for cohort studies with time-to-event-data is based upon survival techniques. In survival analysis the variable under observation becomes time to event (death from lung cancer or censoring). In our analyses, we considered as censored observations those individuals who I I I I I I I

I I I I I I I I I I I I I I I I i I 31 remained, alive at the end of t~e study period (i.e., seven years of follow-up), as well as those with truncated follow-up or who died from causes other than lung cancer. Failure times are computed for all individuals to date of death of subjects who died from lung cancer, our event of interest, or from the remaining causes of death, and time to elnd of the study for all others. As with any observational study, cohort studies are subject to potential confounding. In cohort studies, stratification and multivariate analyses can be I . used to control for confounding. In addition, statistical modeling (i.e., proportional hazards (Cox 1972) regression model or Poisson regression) can be used to estimate the ratio of incidence or mortality rates. Cox regression analysis can provide estimates of the effect of both continuous and discrete variables and for time varying covariates 0~reslow I987). Cox proportional hazard modeling was the primary analytic method used in this study. 3.4. Sub Cohort of Interelst: Definition and Source Population As mentioned above, the study population is restricted to non-institutionalized individuals 30 years and older of households in which at least one household member was 45 years old.I This study will concentrate on nonsmoking participants and their spouses, after applying the inclusion and exclusion criteria described below. A nonsmoker in CPS II is defined as someone who never smoked cigarettes, pipes or cigars or who smoked or smoked less than one of these tobacco products a day ~or one year's time. The distribution of smoking habits in the CPS II at time of enrollment is shown in Table 3. 3.5. Eligibility a) Inclusion Criteria We will include: a) Never smokers. b) Both men and women.

32 ! ! I c) Persons of 30 years and more. d) Persons of all races. e) Period of follow-up: September l, 1982, through December 31, 1989. b) Exclusion Criteria We will exclude from analysis the following persons: a) Current and former active smokers. b) Persons with incomplete or unclassifiable data on smoking habits. c) Persons who had cancer (except non-malignant melanoma skin cancer) at the time of the interview, or whose cancer status was unknown. d) For the analyses of self-reported ETS exposure, we will exclude nonsmoking participants with unelassifiable information on self-reported ETS exposure in any of the following settings: home, work or elsewhere. d) For the analyses of spousal ETS exposure, we will also exclude nonsmoking participants whose spouses are not in the study. e) persons whose spouses have incomplete or unclassifiable data on smoking habits. f) Analyses of intensity, duration and a combined measure of intensity and duration of ETS from spousal smoking will be restricted to cigarette smoking spouses (current and former) with complete data, and who were married only once (both the nonsmokers and their spouses) at time of interview, and who had complete information on age at fast marriage (both for the nonsmokers and their spouses). The analyses of self-reported ETS exposure include 392,226 subjects and 362 deaths from lung cancer. Three people, two women and one men, died shortly after enrollment, and thus did not contribute person-time, and hence were excluded from analyses, bringing the number of subjects down to 392,223 subjects and 362 deaths for most analyses using person-time. The cohort of nonsmokers for ETS from spousal smoking includes 314,108 participants and 265 lung cancer deaths. One of those persons who died promptly after enrollment was a nonsmoking husband and thus did not contribute person-time either, and hence was excIuded from analyses: for analyses based upon person- i I I | i I I i

33 I time data there were 314,107 subjects and 265 lung cancer deaths. Table 4 gives the details of the application of the exclusion criteria to select the major anaiytic cohorts. Another subset of nonsmokers was used as analytic cohort for dose-response analyses of cigarette smoking of the spouses. The time they were married to spouses was estimated, to assess the effect of this variable as well as that of the pack-ycaxs smoked during r~arriagc by smoking spouses. Therefore, wc excluded those spouses married more than once, since the information available on age at marriage in the CPS II questionnaire referred to age at Rrst marriage. In addition to these missing values, there wcrc also missing values (i.e., blanks) for the number of times marric~. Figure 4 shows the sequential application of theses exclusion criteria for analyses of dose-response of ETS from spousal smoking. 3.6. Variables Status The vital status as of December 31, 1989 is a variable assumed one of the following values: I) alive, 2) dead, 3) those who had follow-up truncated on September 1, 1988. The length of follow-up was the difference between date of entry, and date of follow-up truncation (i.e., September 1, 1988), date of death, or December 3 I, 1989, otherwise. Informative events were deaths from lung cancer; if subjects died from other causes, they were censored observations, as were losses to follow-up and subjects alive at the end of the follow-up period. I

34 Table 3. Smoking habits at time of enrollment* CPS II population at,cording to gender of the Smoldng Habits Males Females Total Never smoked regularly 127,165 (25.0) 355,519 (52.6) 482, 684 (40.7) Current cigarette smokers I05,954 (20.8) 135,092 (20.0) 241,046 (20.3) Former cigarette smokers 157,734 (31.0) 138,957 (20.5) 296,691(25.0) Current pipe/cigar and I4,120 (2.8) - 14,120 (1.2) cigarette Pipe/cigar smokers never 22,529 (4.4) - 22,529 (1.9) smoked cigarette Ex-cigarette, 34,649 (6.8) - 34,649 (2.9) ex-pipe/eigar Ex-pipe/eigar, current 19,031 (3.7) - 19,03 l (1.6) cigarette Ex-cigarette, current 11.272 (2.2) - 11,272 (0.9) pipe/cigar Uncertain whether current - 12,822 (1.9) 12,822 (1.1) or former cigarette smoker _Un~classifia..ble ...... 16.140 (..3.2) ...... 34,140.(5.0) .... 50,280 f4,.2) ,. Total 508,594 (100.0) 676,530 (I00.0) 1,185,124 f100.0) So~7~: ACS~ CPS ii documentation' codeboold" ......... * 18 study participants died shortly after enrollment and did not contribute person-time, and four had less than 28 years of age at enrollment. 0 I I I I I

35 Table 4 Number of CPS II persons and deaths from hmg cancer (ICD-9 162) at baseline and nmnber of eligible for analyses of self-reported and spousal ETS exposure Women Men Total ... Persons (%) Deaths (%) Persons (%) Deaths (%) Persons (%) Deaths (%) Total Cohort 676,530 (I00.0) 2,686 (100.0) 508,594 (100.0) 5,470 (100.0) 1,185,124 (100.0) 8,156 (!130.0) Exchlsions I ) Ever smoked 286,871 (42.4) 2,190 (8 ! .5) 365,289 (71.8) 5,174 (94.6) 652,160 (55.0) 7,364 (90.3) 2)Unclassifiable smoking 34, ! 40 (5.0) 95 (3.5) 16,140 (3.2) 137 (2.5) 50,280 (4.2) 232 (2.8) 3) With cancer at baseline: a) lung 154 (0.0) 38 (1.4) 28 (0.0) 10 (0.2) 182 (0.0) 48 (0.6) b) other* and missing 31,172 (4.6) 74 (2.7) 6,014 (I.2) 22 (0.4) 37,186 (3.1) 96 (I.2) data on cancer at interview 4 a) Self- reported ETS exposure 42,655 (6.3) 43 (I .6) 10,435 (2.0) I I (0.2) 53,090 (4.5) 54 (0.7) unclassifiable§ For the spousal cohorl: 4 b) Spouse not .in CPS II w 103,774 (15-.~) 113 f4.2) 15,510 w (3.0) 18-- (0.3) 1"I9,284 (10.1)-" 131 (I.b~ 5) Spouse had unclassifiable 7,265 ( !. I) 12 (0.4) 4,659 (0.9) 8 (0.1) I 1,924 (1.0) 20 (0.2) active smoking (21.8) 116 (2.1) 392,226 (33.1) 362 (4.4) (19.8) 101 (I.8) 314,108 (26.5) 265 (3.2) Analytic cohort for Sel f-reported 281,538 (41.6) 246 (9.2) I 10,688 . ETS exposure Analytic cohort for Spousal ETS 213,154 (31.5) 164 (6.1) 100,954 . ..exposure ~' All other except non-melanoma skin cancer. § ttours per day coded as unclassifiable for any of home, work or other 2063620573

36 I Figure 4. Exclusions for analyses on dose-response of ETS from spousal cigarette smoking Exclusions: Married to ever pipe/cigar smokers and with incomplete cigarette smoking data 55,609 nonsmokers and 49 deaths 65,914 Married more nonsmokers than once or and 63 deaths missing data Missing age 44,381 at marriage nonsmokers and 26 deaths 3 I4,108 nonsmokers [ and 265 deaths ~258,499 nonsmokers and 2~death~ ~,.~192,585 nonsmokers arid 15~eaths 148,204 nonsmokers and 127 deaths* *Population of 148,204 comprises study group for tables 23, 38-40 Main Outcome Death from lung cancer: Subject reported as deceased as of December 31, 1989, from cancer of the trachea, bronchus and lung (ICD-9 codes 162.0 to 162.9) as the underlying cause of death. 0 0~ 0 I I i I ,I I I I I I I

I I I- I I I I I I I I I I I I I I 37 Exposure Variables Main Exposure Variable (ETS) i. Self-reported ETS exposure The average number of hours per day a person reported being presently exposed to ETS at home, work and other places will be used as one exposure variable (i.e., self-reported ETS exposure at time of enrollment). We will use the number of hours a person is exposed at home, work and other places, as well as the sum of the exposures in the three settings, as exposure variables. Dichotomous ETS exposures will be examined (none versus any), and we will then examine the number of hours of exposed individuals, ~ouped in tertilcs, and then treated as ordinal variables for dose-response hypothesis testing. ii. Spousal Smoking Habits A second source of information on ETS exposure involves linking non-smokers with the active smoking habits of their spouses. The questions on active smoking previously described, plus the information on times married and age at marriage, and age at interview (both for the index subjects and their spouses), and age at uptake and cessation of smoking for smoking spouses, were used to estimate the intensity and duration of ETS exposure from spousal smoking. Smoking status (ever and never) as well as cigarettes usually smoked per day by current and former smokers will be considered. Pack-years of cigarette exposure from the spouse will also be calculated by multiplying the number of cigarettes smoked per day, by the number of years the spouse smoked cigarettes while married to the study subjects. For the quantitative analyses dealing with intensity, duration and the combination of these two dimensions of ETS exposure from spousal smoking, we will restrict the analysis to subjects with valid information on these variables above mentioned, as needed to estimate time in marriage exposed to ETS from spousal smoking. These will be the other main

38 I I I exposure variables for a subset of the spousal cohort described above. Pack- years of cigarette smoking during marriage were ~ouped by quintiles of the distribution of ETS exposure from spousal cigarette smoking. Covariates Potential confounders and effect modifiers included in the analyses were : l) age, 2) gender, 3) socioeconomic status, as measured by schooling, and race; 4) exposure to asbestos, 5) frequency of consumption of six groups of fruits/juices and vegetables, as major sources of carotenoids, 6) total dietary fat as nutrient index, and 7) a history of tuberculosis and other chronic diseases of the lung. All these variables were examined as independent risk factors and controlled for in the analyses. A complete discussion of potential confounders of the ETS- lung cancer link can be found elsewhere (Buffer 1990, Mantel 1992). The analysis of these covariates in the CPS II cohort is complicated by missing data on some of these variables. As will be discussed below for ETS data in CPS II (Sections 3.8 and Chapter 5), a large proportion of CPS II participants left blank the relevant spaces provided in the questionnaire. CPS II participants were not instructed to record zeros for no consumption/exposure. Regarding demographic data (i.e., age, and gender) and schooling and race as proxies of socio-economic status, there are few subjects with missing data. For two of these four covariates: 'race' (1,393 subjects, or 0.4 percent) and schooling (5,413 subjects or 1.4 percent), subjects will be treated as a separate strata in multivariate analyses, when blocking on them. Only when obtaining estimates for covariates with missing data (i.e., included in the model instead of blocking for them), an indicator variable of missing data will be set up, and hence we did not block for them (Table 30). However, there are many more subjects with missing or blank data on items in the food frequency, history of chronic lung diseases and occupational history sections, from which the other set of covariates was derived. The approach we used to deal with these missing data for food frequency is similar to that followed by Thun et al. in a nested case- control study of colon cancer in CPS II (Thun et al., 1992). Missing values I I I I I I I I I / 0 0"t 0",

39 I were considered to represent infrequent consumption and a 0.25 times per week value was imputed. Thun et al. validated their approach by comparing the prevalence of reported consumption of several food items in CPS 1-r with data from the National Health and N~atrition Examination Survey epidemiologic study. The consumption patterns in the two datasets were similar when missing values were grouped with the categories of 0 or <1 time a week in CPS 1I. Thus, for analytic purposes we let blanks be zeros and assigned a frequency of once a month (i.e., 0.25 in a weeklYlscale ranging from 0, none, to 7, everyday) to answers of < 1 time per week. Our approach also follows that of Thun et al., in excluding from the multivariate analyses of diet, persons who completed fewer than five food items or who left blank an entire column of the questionnaire. For the purposes of adjusting for ~tis covariate and providing an estimate of other variables such as ETS, these subjects with incomplete data in the diet section were treated as having missing values, but considered valid values of these covariates and allowed to form strata as such and will not be excluded. Following the same rationale, blanks in data on medical history of tuberculosis and chronic obstructive pulmonary disease and occupational exposures were considered to represent a negative history of such disease and no exposure, respectively. As an alternative,source to self-reported occupational exposure to asbestos, any mention of having ever held any of occupations likely to involve exposure to asbestos (i.e., shipbuilders, pipefitters, as high-dose exposed, and plumber, construction, duckworker, autorepair, and electrician, as low-dose), or possibly exposed (i.e., janitor, railroadworker, foreman, machinist, painter, ! assembler, welder, miner, sewer, factory worker, firemen, engineers, steel mill workers, aides, laborers, refinery workers, and military) was used to compare the mortality experience with that of those who had occupations unlikely exposed. We followed the apl~roach of Hinds et al. (1985) to rate these trades: two raters independently classified the list of occupational codes in CPS II (VMC and Howard Frumkin, M.D., Dr. P.H.), agreeing in most of the cases. The final list was completed with the instructions of Dr. Frumkin. The classification here presented closely agrees with that of Hinds et at. (I985), under the I circumstances of sketchy data on occupations available in CPS 11.

4O I I I The same variable derived in a recently published analysis of CPS IT data (Thun et al., 1992) to assess fat intake from the food frequency section will be used. Briefly, this variable represents a nutrient index calculated for each person by summing the products of consumption frequency of each food item by the fat content of a medium-portion size for that food (specific to age and sex as estimated for US adults from the NHANES TT survey). Total fat consumption was divided into quintiles, and CPS II nonsmokers in lowest quintile were the referent group. As for total frequency of foods and vegetables containing carotenoids, subjects who insufficiently filled the CPS II questionnaire in the diet section were coded as missing for total amount of dietary fat. Denominator Information Person-years denominators were summed over five-year age intervals for the time each person was observed. This procedure provided the appropriate denominators for mortality rates. Person-years accumulated during follow-up were stratified by specific characteristics (i.e., exposure variables and covariates), such as age and smoking history of cohabitees. A_data step 'macro' procedure for the Statistical Analysis System was developed by the ACS Division of Epidemiology and Statistics that provided person-year stratification. The follow-up period of a study subject was subdivided into segments of months and, for each segment, age was evaluated at its midpoint. 3.7. Validation of ETS exposure data Self-reported data on current exposure to ETS were validated using two approaches: Because the CPS II questionnaires did not require respondents to complete all fields, many questionnaires contained blanks. To detetanine whether these blanks should be considered negative responses or incomplete question.naires, I I I I I I I I I I

I I- I I I I I I I I I I I I I I 41 we com~re~ CVS-L~ data on E[$ e×posure"at home with <iata Crom (:he [~8~ Nationm Health Interview Survey (MMWR 1992, NCHS 1988). No previous population based surveys inquired about ETS exposure. The 1988 NHIS included a set of questions for adults about their lifetime working status and their work experience in the year before the interview. The I988 NHIS included information about ETS exposure for 44,233 respondents, based on the questions: "Do you live with a smoker?", and "Do they smoke at home?". Comparisons were made using direct standardization for age, race and gender, and taking the weights from the 1980 US Population (US Census Bureau 1983). A second validation study involved a comparison within CPS II, comparing self- reported ETS exposure with the smoking status of cohabitees and spouses enrolled in the CPS II cohort. Persons living with or married to nonsmokers or former smokers should be less likely to report any current exposure to ETS at home than persons living with or married to current smokers. We tested the agreement between these two independent measures of exposure to ETS. Thus, the number of current smokers among household members in CPS II was estimated and compared with the self-reported number of hours of ETS exposure at home. The smoking status of spouses was also compared with the self-reported number of hours of ETS exposure at home (e.g., smoking status of spouse versus number of hours exposed to ETS, and packs of cigarettes smoked by current cigarette smoking husbands). The agreement correcting for chance was measured using the k statistic (Fleiss 1981). For the most simple case of the agreement between two observers, a two by two table analysis is displayed as illustration of the method: Observer I Observer 2 Present Absent Total Present p I I p 12 p 1- Absent p21 p22 p2- Total p- 1 p.2 n where Pii are expressed as fractions of n (i.e., total sample size), and. denote marginals. O

42 the agreement by chance Pe is given by Pe = (Pl.P. 1 + P2.P.2) and the observed agreement Po by k Po = ~Z1.= where k denotes the observers (two in our example). In our example:/~ = pl 1 + p22 [(pl o*p ° I) + (p2 • *p ° 2)] n l pll+p22 n The standard error of k, letting Pij be the proportion of subjects assigned to category i by rater I and category j by rater 2, SE(~)= ~-ee4g where k a = i~=l Pii[l - (Pi. + p.i)(l -/~)]2 b=(1-k)2X X Pij(P.i +pj.)2 i#j and c=[~-pe(1-~)]2 (Cfr. Brilliant et ai.,1983). Interpretation of estimates of k follows the criterion outlined by Landis and Koch: "values greater then 0.75 or so may be taken to represent excellent agreement beyond chance, values below 0.40 or so may be taken to represent poor agreement beyond chance, and values between 0.40 and 0.75 .may be taken to represent fair or good agreement beyond chance" (Fleiss 198 I). I I ! I I I I I

43 I I I I I I I I I I I I I I I poor agreement beyond chance, and values between 0.40 and 0.75 may be taken to represent fair or good agreement beyond chance" (Fleiss 198 I). 3.8. Exposure Criteria Used In Analysis i. Self-reported ETS Exposure Persons with blank spaces in the questionnaire for ETS at home, work and other places were considered unexposed (i.e., 0 hours of exposure). Persons with unclassifiable information on ETS exposure were excluded. The three fie[ds (hours of ETS exposure at home, work and other places) were added to obtain a cumulative exposure variable. ii. Spousal ETS Exposure Most published epidemiologic studies have relied on spousal smoking history, rather than on self-reported exposure. Indeed, as shown in the recta-analysis presented above, the strongest evidence of a causal relationship derives from spousal smoking. Spouses generally have a closer and longer relationship with the study subjects than do other adult household members. Therefore, we will use information on the smoking status of spouses, number of cigarettes, pipes and cigars smoked and for spouse-pairs married once in their lifetime, we computed time in marriage nonsmokers were exposed to ETS from spousal cigarette smoking as described before combined with the informationbn the quantity usually smoked by spouses to estimate pack-years. 3.9. Validation of Information from Death Certificates Metastatic cancers to the lung may comprise a larger proportion of "lung cancers" in non-smokers than among smokers. A number of authors have studied the accuracy of cancer death certificates by comparing the specified underlying cause of death to autopsy diagnosis and more specific hospital and pathologic data. In one of those studies, using data from eight of the nine areas included in the Third 0 O~ 0", 0 O~

National Cancer Survey (TNCS), the underlying cause of cancer deaths as determined from death certificates was compared to the hospital diagnoses for 48,826 resident cases of single primary cancers. The death certificate diagnosis was confirmed by the hospital diagnosis in 9,560 (93.9%) out of 10,178 lung cancer deaths (Percy 1981). To validate the use of information on lung cancer diagnoses from death certificates, we conducted a validation study. In particular, we assessed whether deaths coded as "lung cancer" in nonsmokers were truly primary lung cancer or metastatic from other sites. For 30 deaths for which lung cancer was considered the underlying cause of death in CPS ]I nonsmoking participants who resided in SEER registry areas, we compared SEER diagnosis with underlying cause of death on death certificates. These 30 deaths represent 9.7 percent of all deaths from lung cancer as of August, 1988, among never smokers free of cancer at the beginning of the follow-up (i.e., 296 deaths of "incident cases" that had occurred as of August 1988). The NCI-SEER Program cancer registries cover approximately 9.5 percent of the US population (NCI SEER 1991). Finally, we reviewed each one of the death certificates in the analytic cohorts, and checked for inconsistencies in the selection or coding of the.underlying cause of death. In doing so, our assumption is that the most critical parameter for the purpose of the study validity is in its specificity rather than sensitivity (Kleinbaum 1982). 3.10. Statistical Analysis Outline of the Analytic Approach The analytic approach to be used is outlined in Figure 5. After checking and editing the main exposure variables and covariates described below, the analyses followed these steps • 0 0 ! I I I (? I I I I I I

| I I I I I I I I I I I I I I I I I 45 I) For the self-reported ETS cohort: a) Simple and stratified analyses using rate ratios of any versus no ETS exposure, followed by multivariate analysis using Cox regression to adjust for potential confounding. b) Rate ratios of hours of ETS exposure per day (summed over the three settings: I home, work and other places) grouped by tertiles, and using this varibles as categorical fast, and then as ordinal in Cox regression analysis, eontrolling for potential confounders. 2) Spousal ETS cohort: a) Simple and stratified analysis using rate ratios of the smoking status and type of smoking habits of spouses (ever versus never, any cigarette, pipe/cigar versus I never), followed by multivariate analysis using Cox regression to control for potential confounders. Additionally, the amount of cigarette smoked during marriage was ~ouped into quintiles and compared using the rates of nonsmoking participants married to nonsmoking spouses as the reference. i Also we assessed the strength of the association of potential confounders with the risk of lung cancer, as well as the distribution of covariates among the exposed and unexposed to ETS (both self-reported and from spousal smoking). 0 O~ O~

46 i Figure 5. Outline of Analytic Approach Entire cohort t.2 million Self-Reported ETS 392,226 subjects* Analysis on self reported number of hours of ETS Rate ratio analysis and Cox regression for grouped and ordinal dam Excluding smokers, persons with cancer at the start of f/u, etc. Spousal ET$ 314,108 sllbjeets** Analysis on smoking history of spouses (status and quantity) Hypoth~es 1) Test overall association Bivariate and multivariate Rate ratio analysis analysis (adjusted for and Cox regression for confounders) grouped and ordinal data ~ 2) Concomitant variation t48,21M I subjects*** Bivariate and multivariate analysis (adjusted for For duration and confounders) intensity*duration of ETS from spousal cigarette smoking *Tables 25-33 and 41 **Tables 34-37 ***Tables 38-40 Using the methods described below under the sections of simple and stratified analyses and Cox regression for both the self-reported ETS and spouse-pairs cohorts, the null hypothesis of no association between exposure to ETS and lung cancer was tested comparing the rates of lung cancer among nonsmokers according to self-reported hours of exposure to ETS and the smoking status of o o I I I I l I I I I i

I I- I I I I I I I I I I I I I I I 47 their spouses, respectively. Age-adjusted comparisons was added in ~hc simple and stratified analyses, and Cox regression. The covariatcs (socioeconomic level: 'race' and schooling, gender, age in 12 five-year groups, history of chronic lung disease, and occupational exposure to asbestos) were blocked for in stratified Cox analyses with the ETS main exposure variables in the models, to make fewer assumptions regarding the proportionality of the hazard rates. These variables were entered together into the regression models with ETS variables (self-reported and spousal), [and s~lected because of a priori knowledge. Intensity, duration, and a combination of these two dimensions of tobacco smoking during marriage of non-smokers (i.e., study subjects) to smoking spouses (i.e., the exposure) was examined as continuous and grouped data, to test the hypothesis of a dose ~sponse relationship. Thus, l) the number of cigarettes smoked per day by current and former smokers with complete data, 2) the years of ETS from spousal smoking during marriage, and 3) pack-years smoked by cigarette smokers during marriage were used as exposure variables for these analyses. Exploratory Data Analysis First, the exploratory analysis lof the data provided information on the simple and cumulative frequencies, measures of central location and dispersion, identification of extreme and missing observations, and subsets of variables and records of interest. ETS variables were explored along with other variables mentioned in the 'Exposure information at baseline' section, and grouped into quintiies or tertiles. Simple and Stratified Analysis The parameter of interest in this study is the ratio of mortality rates for selected characteristics of the study participants. For large populations, like CPS 13, and short periods of follow-up, the risk of mortality approximates mortality rates. 0 O~ O~ 0 O~

48 We used the binomial distribution in the statistical analyses of simple 2x2 tables and stratified analyses. Under the null hypothesis, the number of exposed cases a, is unrelated to that among the unexposed, using the theorem of conditional probability, and given that the total number of cases M1 is f'txed, a is a variable that follows a binomial distribution (Breslow 1987). Also, it has been shown that given the distribution of one of those binomial variates conditional on the total number of cases, a follows the hypergeometric distribution; and if the conditions MIT = 0,@- ~ 0are met, the limit is a Poisson distribution with parameter MtNI (Miettinen 1985). T Exact binomial 95 percent confidence intervals around rate ratios were calculated. Simple analyses were followed by a stratified analyses. A basic feature of the method used to estimate age-time-specific mortality rates consists of determining for each subject the amount of follow-up time contributed to a given age times calendar period category and to sum up those contributions for all the members of the CPS II cohort to obtain the total number of person-years of observation in a given category. Since ratios of age-specific cancer~ incidence are more constant, the measure of choice was ratio of rates rather than rate differences. Age-adjusted rates were calculated using the entire CPS 1I person-years population as standard. The formula for a standardized rate is: SR _ Z W iRi Z,wi Weighted averages of stratum-specific effect measure estimates were obtained using as weights the product of the weight from the standard and the rate among the unexposed: I I ! I I I I ,I I I I I I I

I I I I I I I I I I I I I I I I 49 Precision-based variance estimate for this expression has two terms using approximate variance estimates of the rates to substitute then in the formula: Var[ln(SRR)I = I (2WiRli)2 (~, wiRoi)2 I Comparison groups were each one of the self-reported ETS and spousal smoking categories taking the unexposed categories (i.e., never smokers) as referent. Categorical variables were formed using the unexposed as referent and dividing exposed into tertiles or quartile. For the stratified analysis the Mantel-Haenszel method (Mantel 1959) was used. This method uses as weights the contribution of unexposed cases times the number of exposed person-years to the total of each stratum: it is a simple noniterative estimator for a uniform rate ratio and is nearly as efficient as the maximum lildelihood estimator (Rothman, 1986): i /'i IDq(m-h) = ~biNli/~T. i where ai and bi are exposed and unexposed cases, Nli and Noi are exposed and unexposed person-time Idenorrdnators, and Ti are the totals for the i stratum. Rothman (1986) reviewed variance estimators of the above point estimator of the Mantel Haenszel approach. A stable formula for the variance that considers each ai to be an independent binomial vadate conditional on Nli is: 0 O~ IX] 0

50 iV MliNliNOi y. i=l T2. Var[In(ZDRmh)]-. ~ aiN.oi. ~ biNli i=1 Ti i=1 Ti Both the Mantel-Haenzsel and the maximum likelihood summary rate ratio estimators were computed using a "rapidly converging network" algorithm (Guess and Thomas 1990), which in turn uses an F-distribution algorithm developed by Brownlee as used in progrzm~ #12 and #15 of Rothman and Boice (Rothman 1982). A program that uses these algorithms and developed by Simons, Campos-Filho and Nechi (IDR-E) which provides mid-p values for exact binomial confidence limits, was used. Confounding and effect modification were assessed following standard criteria (Miettinen 1981, Greenland 1989), and the stratified analysis led to select variables for statistical modeling. After reviewing the published literature comprising more than 30 reports Of epidemiologic studies which found no evidence of confounding, it was anticipated that confounding in this study was unlikely to occur. Confounding by age, marital status, and education (as an indicator of socioeconomic status) was assessed. Even if no change in esdmate by these potential confounders was found, we obtained and reported, at least, age-adjusted rate ratios, or age-gender adjusted RR when appropriate given that age is the major determinant of the risk of lung cancer among nonsmokers. Assessment of confounding by other well established risk factors such as SES (i.e., 'race', and schooling), dietary intake of foods containing carotenoids, dietary fat, and asbestos exposure was also conducted, because we assess the" ETS-lung cancer hypothesis only when other factors known a priori to be causally related to lung cancer are taken into account. Effect modification was predicted to occur by asbestos occupational exposure. It is known that active l I I I I I | I i I

I I I ! I I I I I I I I I I I I I 51 smoking acts syneroistically with asbestos exposure to account for excess risks among those exposed to both asbestos and active smoking (Selikoff 1968). I Multivariate Analysis Multivariate analyses using Cox regression analysis (Cox I972) were used as the major analytic tool in this research. The general form of the stratified model which uses a partial likelihood function is: I %g(t,X) = ZOg(t)eblxl +b2x2 +...+bpXp where t represents a continuous variable (i.e., length of follow-up, age); ~(t) is I the mortality rate of persons with specified values of variables XI, X2, ...Xp; and ~o (t) is the baseline mortality rate (i.e., of unexposed persons) at the t level of the continuous variable, and the g subscript indicates the g-th stratum of categories of the covariates in ~ae model. I The choice of the model is based upon the type of data. In order to obtain the greatest control over confounding by age and other covariates, the proportional hazards model takes into account the contribution of the set of person-time "at risk" and provides adjustmer~t for covariates to simultaneously estimate their effects or to block those covariates for which the proportional hazard assumption might not hold. In addition, this model has fewer assumptions (i.e., assumes no parametric distribution, on.ly that the rates have the same ratio over time, a multiplicative model takes ptace, and that the occurrence of disease in each subject is independent of the occurrence of disease in other subjects). Hypothesis testing was carried out using the likelihood ratio statistic [-21nL / reduced model minus - 21nL hall model], via maximum likelihood estimation procedures available in the PHR.EG Procedure of SAS (SAS 1991). A formal test of heterogeneity was provided by the likelihood ratio test for fitting the proportional hazards model. The change in estimate criterion and allowing for a ~0,

52 I 1 I priori knowledge of potential confounders (i.e., by age even if it had not shown up as confounder in the data) (Greenland 1989) were used in model building. Concomitant variation in the stratified analysis step was assessed contrasting the rates of lung cancer among ETS unexposed non-smokers to k categories of ETS exposed non-smokers. Ordinai variables were created from categories of dummy variables to test the hypothesis of increasing rates by increasing levels of exposure to ETS using the likelihood ratio test. We treated k number of categories of cumulative exposure (i. e., k categories of number of hours exposed to ETS, or pack-years of cigarettes smoked by spouses), as continuous variables. Adjustment for covariates was allowed in testing this hypothesis by blocking for them. Regression diagr~ostics used include plotting survival curves [log -log (S l(t)) and log -log(S0(0] and checked for a pattern of parallelism (a constant ratio). For most analyses the estimates were obtained by blocking for them, rather than including them in the model However, when estimates were obtained for the covadates, all of them along with the main exposure were included in the model. Kaplan-Meier survival estimates were computed for the main exposure variables, as well as the covariates and the above mentioned graphic approach was used to check for the proportional hazard assumption. 3.11. Sample Size and Power Considerations The statistical power attained by the sample size of this study to detect different values of the rate ratio, including the point estimates from this study, was computed using the following estimator that assumes the rate ratio is a binomial parameter (Breslow 1987): 0 O~ 01 -0 0 I I I I I I I I

53 I I I I I I I I I I I I I I t-3=t-O(A)= where XA is the most extreme value in the acceptance zone under the the null hypothesis. One way of estimatin~ XA is by using the beta distribution with parameters 1- (z and tho expected number of exposed and unexposed cams under the null.

54 Chapter 4: Comparisons of Demographics and Smoking Habits in the US, CPS II, and the Study Populations Rationale A comparison of the 1980 US population with the CPS II population and specifically with the two analytic cohorts (i.e., 1) self-reported ETS and 2) spousal ETS by gender, age, race, occupation, education, geographic residence and smoking habits, is presented below. We used the population figures from the 1980 US Census as standard for age-adjustment, unless specified otherwise, because it was the Census closer in time to the cohort at the time of enrollment. Therefore, we excluded for the purpose of these comparisons those CPS II participants who resided in Puerto Rico, since they were not included in the 1980 US Census. Comparisons with the 1983 National Health Interview Survey (US DHHS Surgeon Genera/1989) figures are a/so presented. These comparisons lend a general perspective to better understand the analytic cohorts, and particularly to generate a profile of the demographics and smoking habits of the subjects in the study cohorts. Race Demographic information in the US is available by 'race'. Race is a proxy of socioeconomic status and was used here for the purpose of demographic comparisons. Twelve percent (or 26 million) in the US are blacks. In CPS II they represent 4.4 percent (or 52,038) of the participants. For these masons, further comparisons of demographics were restricted to whites. Gender and Age Structure The ratio of males to females (or gender ratio) in CPS II is considerably lower (0.75) than that among persons 30 years and older in the 1980 US Census 0 I I I I I / I

I I I I- I I I I I I I I I I I I I I 55 (0.88). Participants in this larg~ cohort were more likely to be in their 50's 60's at enrollment (Tables 5 and 6). Nonsmoking men and women (i.e., in our analytic cohorts) did not differ in their age distribution from the entire cohort (median 57 years in both groups). I Table 5. Comparison of age distributions of white males 30 years and over, in the US population in 1980, with CPS II participants*, .................. andI analytic c, ohorts* ..... Men Age 1980 Entire Self- Husbands Group Census % CPS II % reported % (spousal % 30-34 7,386,562 16.1 7,610 1.6 3,078 3.0 1,126 1.2 35-39 5,848,891 12.7 9,270 1.9 2,890 2.8 1,875 2.0 40-44 4,862,473 10.6 15,052 3.2 3,890 3.7 3,.286 3.4 45-49 4,616,347 10.1 6~,776 14.4 l 7,079 16.4 16,003 16.8 50-54 4,925,489 I0.7 87,030 18.2 19,14I 18.4 18,480 I9.4 55-59 4,877,635 10.6 91,236 19.l 17,647 16.9 16,893 17.7 60-64 4,199,446 9.1 79,344 16.6 15,804 15.2 15,306 16.0 65-69 3,470,295 7.6 58,162 12.2 11,861 11.4 11,406 11.9 70-74 2,565,929 5.6 35,487 7.4 7,069 6.8 6,534 6.8 75-79 1,652,668 3.6 17,045 3.6 3,687 3.5 3,206 3.4 80-84 918,166 2.0 5,909 1.2 1,361 1.3 1,029 1.1 85+ 603,663 1.3 ~.419 0.5 624 0.6 330 0.3 Total 45,927,564 I00 477.340 i00 104,131 100 95,474 100 *Excludes CPS II participants who resided in Puerto Rico

56 Table years and over, in the US population in 1980, with CPS II I I I I I ! I ! ! ! 6. Comparison of age distributions of white females 30 participants*, and analytic cohorts* Women Age 1980 Census Entire Self- Wives Group % CPS H % reported % (spousal % ETS ETS) 30-34 7,411,223 14.2 I 1,764 1.9 5,591 2. I 2,971 1.5 35-39 5,949,670 11.4 18,831 3.0 7,579 2.9 5,753 2.8 40-44 4,981,237 9.5 44,595 7.1 18,241 6.9 16,858 8.3 45-49 4,807,473 9.2 91,972 14.7 37,349 14.2 34,006 16.8 50-54 5,249,428 I0.0 106,175 17.0 43,434 16.5 38,805 19.2 55-59 5,409,320 I0.3 107,900 17.2 43,756 16.7 38,098 18.8 60-64 4,826,403 9.2 92,102 14.7 38,274 14.6 30,949 15.3 65-69 4,344,316 8.3 68,889 11.0 28,367 10.8 19,637 9.7 70-74 3,562,454 6.8 44,568 7.1 19,731 7.5 10,295 5.1 75-79 2,667,233 5. I 23,892 3.8 11,736 4.5 3,866 1.9 80.84 1,756,793 3.4 9,916 1.6 5,366 2.0 881 0.4 85+ 1,400,053 2.7 5,350 0.9 3, I65 1.2 160 0.1 Total 52,365,603 100 625.954 I00 262,589 100.0 202,279 100 *Excludes CPS rr participants who resided in Puerto Rico Occupation The types of occupations presently held by CPS H employed white participants were categorized into white and blue collar occupations. Managerial and professional specialty occupations, technicians and related support occupations, sales occupations, and administrative support occupations including clerical represented white collar occupations. Precision production, craft, and repair occupations, operators, fabricators, and laborers were classified as blue collar occupations. For these comparisons we excluded subjects with the following occupational codes in CPS II: housewives, disabled, retired, and subjects with none or unspecified data on occupations. I i I I

I I I I- I I I I I I I I I I I I I I 57 CPS II participants were more likely to be engaged in white collar occupations (Table 7). White women in CPS II were more likely to hold white collar jobs than white men, in a higher proportion than their counterparts in the entire US population do. Nonsmokers di~ not differ from the entire cohort with respect to their occupations. Table 7. Comparison of occupations of employed white persons 30 years and over, in the US population in 1980, in CPS II participant*, and analytic cohorts* a. Men 1980 Entire Type of Census CPS 17 Jobs (%) f (%) White Collar 18,I65,788 200,612 (55.8) (73.7) Blue Collar 14,409,714 1 71,718 (44.2) (26.3) Total 32,575,502 272,330 (100.0) (100.0) Self-reported Husbands ETS (spousal (%) ETS) (%) 47,889 43,901 (74.3) (73.7) 16,573 I5,684 (25.7) (26.3) 64,462 59,585 (100.o) (IOO.O) b. Women 1980 Entire Self-reported Wives Type of Census CPS II ETS (spousal Jobs (%) I (%) (%) ETS) (%) White Collar 18,464,642 221,093 91,700 70,404 (84.8) (94.6) (94.3) (94.5) Blue Collar 3,299,972 I 12,553 5,518 (5.7) 4,137 (5.5) (15.2) (5.4) Total 21,764,614 233,646 97,218 74,541 (lOO.O) (lOO.O) (ioo.o) (ioo.o) *Excludes CPS II participants who resided in Puerto Rico 0 Ol 01

58 Schooling Nonsmoking CPS IT men and women were more educated than smokers in CPS II as is also true for the rest of the US population, as reflected by their considerably higher rates of college graduates (Table 8). The entire CPS II cohort, after adjustment for age is also more educated than the US population as a whole (28% of college graduates in CPS II women versus 12% in the US populations over 30 years of age). Nonsmoking men in the analytic cohorts (in the cohort for analyses of self-reported ETS and among nonsmoking husbands for analyses of ETS from spousal smoking) were more educated than the rest of the CPS rr men. Table 8. Comparison of the proportion (%) of college graduates among whites in the US population in 1980, CPS II participants§, and analytic cohorts§ I- Men -I I- Women -I Age US CPS II SRETS Hus- US CPS II SRETS Wives • ga~o~ up .... C.ensus ~[ bands Census q[ 30-34 31.5' ' 49.~' 62.5 66.0 21.4 40.4 47.1 43.7 35-39 27.7 48.4 61.1 64.3 17.2 34.9 39.9 37.9 40-44 23.6 43.8 56.7 56.6 13.6 30.8 32.0 31.3 45-49 22.6 46.2 56.9 56.5 11.6 28.9 28.8 28.4 50-54 19.7 43.1 53.1 52.4 10.3 26.6 25.3 25.0 55-59 t 7.0 39.0 47.6 47. I 8.5 23.0 21.7 21.4 60-64 13.2 32.6 40.0 39.8 8.2 21. I 20.3 20.6 65-69 11.4 27.1 32.9 33.2 8.0 20.8 20.0 20.5 70-74 II.I 26.1 30.2 30.9 8.5 22.4 21.1 22.2 75+ 9. I 24.8 27.8 30.0 6.7 21.2 13.0 20.3 Adjusted 20.6 40.4 50.1 51.0 12.0 27.9 28. I 28.2 § Excludes CPS II participants who resided in Puerto Rico ~ SRETS: Serf-reported ETS cohorts * The standards are taken from 1980 US Census race-gender specific populations 0 0"~ O- I I I I i I ! I I I I I I

59 I I I I I I I I I I I I I I I Marital Status As shown in table 9, CPS I1 participants were more likely to be married than the rest of the US population, a fact that may be related to their more affluent status and the way they were enrolled. There were more unmarried women, and particularly single women in CPS II than unmarried men. This difference may be explained by a more active participation of women in recruiting people (i.e., ACS volunteers), whereas the men ~ere more likely to get enrolled in CPS 11 as members of family ~oups. Table 9. Comparison of the proportion (%) of married whites in the US population in 1980, white CPS II ..~ partieipants§~ and lanalytie cohorts.§ Men Women Age group US CPS 1I SRETS US CPS II SRETStJ[ __ Census ~ ~[ Cens~us _ . .................. 30-34 77.4 62.4 60.61 79.4 67.4 67.6 35-39 83.5 80.7 77.2 82.0 82.3 82.4 40-44 85.8 91.3 90.4 82.8 91.4 92.4 45-49 86.6 95.9 95.7 82.2 88.6 90.9 50-54 86.5 96.4 96.5 79.6 87.0 89.1 55-59 86.8 96.8 96.6 75. I 83.3 85.9 I 60-64 86.3 96.7 96.8 ' 67.2 77.0 79.9 65-69 84.2 95.9 96.2 56.2 65.7 68.4 70-74 80.8 94.2 94.3 43.6 51.7 54.3 75-79 74.8 90.8 91.7 30.1 35.3 36.2 80-84 65.3 83.2 83.6 17.9 19.8 19.9 85+ 48.8 62.0 60.8 : 8.3 7.9 7.7 Age 82.8 87.3 86.5 67.9 71.8 73.3 Adjusted* §Excludes CPS II participants who resided in Puerto Rico ~ SRETS: Self-reported ETS cohorts * The standards are taken from the 1980 US Census race-gender specific populations

6O I Residence The distribution of the CPS II by territory in general resembles the distribution of the US population (Table I0). A few States in the South (e.g., Texas, Oklahoma), the Mid-West (e.g., Missouri) and the North-East (e.g., New York) showed a deficit with respect to the distribution of the US population. Two States, Minnesota and Utah, had an outstanding participation rate, reflecting the activities of the ACS Divisions and perhaps the advancement of public health in those communities. Smoking Habits The age-adjusted prevalence of smoking habits in CPS lI and the 1983 HIS is shown in Table 11. Prevalence fi~mares of smoking habits in CPS 1I are similar to those of the US population by 1982. ~0 0 0~ I I ! I I I I I <1 I :~1 I '1 I I

I I I I I I I I I I I I I I I I I I Table State US Alaska 0.2 Arizona 1.2 Arkansas 1.0 Caiifomia 10.5 Colorado 1.3 Connecticut 1.4 Delaware 0.3 D.C. 0.3 Florida 4.3 Georgia 2.4 Hawaii 0.4 Idaho 0.4 Illinois 5.0 Indiana 2.4 Iowa 1.3 Kansas 1.0 Kentucky 1.6 Lousiana 1.9 Maine 0.5 Maryland 1.9 Masss 2.5 Michigan 4. I Minnesota 1.8 Mississippi I. I Missouri 2.2 10. US population in 1980, and CPS II by State of residence cPS rrI Men Wamen 1,7 1.8 0.I 0.I 1.4 1.4 1.3 1.3 8.7 8.9 1.2 1.2 1.7 1.7 0.3 0.3 0.1 0.i 4.8 4.9 2.6 2.6 0.2 0.2 0.6 0.5 5.6 5.6 2.8 2.8 1.5 1.4 1.5 1.5 1.5 1.6 0.9 1.0 0.6 0.6 2.8 2.7 2.0 2.0 3.8 3.7 3.2 3.0 0.9 0.9 1.3 1.3 participants* CPS II State U S Men Women Montana 0.4 " 0.4 0.4 Nebraska 0.7 I. 1 1.0 Nevada 0.4 0.2 0.2 N Hamp. 0.4 0.4 0.4 New Iet~ey 3.3 3.7 3.7 New Mcx 0.6 0.5 0.5 New York 7.8 5.8 6.0 NCarolina 2.6 1.8 1.8 N Dakota 0.3 0.6 0.5 Ohio 4.8 4.5 4.5 Oklahoma 1.3 0.0 0.0 Oregon 1.2 1.3 1.4 Penn 5.2 6.4 6.5 R Island 0.4 0.7 0.7 S Carolina 1.4 1.2 1.3 S Dakota 0.3 0.6 0.5 Tennessee 2.0 2.6 2.7 Texas 6.3 4.6 4.5 Utah 0.6 2.0 1.9 Vermont 0.2 0.2 0.2 Virginia 2.4 2.7 2.8 Washington 1.8 1.8 1.8 W Virginia 0.9 1.0 I. i Wisconsin 2.1 2.7 2.6 Wyoming 0.2 0.2 0.2 *Excludes CPS II participants who resided in Puerto Rico 61 I

Table 11. Age-adjusted prevalence* of current, former, and never cigarette smoking, CPS II and NHIS-83 Current Former Never CPS H HIS CPS HIS CPS HIS II H Men White 30.1 31.1 44.4 43.5 25.5 25.4 Black 42.5 41.8 31.6 32.1 25.9 26. i Women White 20.4 26.0 22.5 19.7 57.1 54.3 Black 26.2 27.4 I5.8 14.4 58.0 58.2 *(Percent). Standard population: CPS H 62 I i I I I I I I I I I I I I I I I

I I I- I I I I I I I I I I I I I I I 63 Chapter 5: Validity and Completeness Information I On the Outcome Variable i. Follow-up Procedures of the A validation study of the CPs II automated follow-up procedure has been conducted previously using the National Death Index (ND1) (Calle, 1993). In a linkage of over 15,000 persons whose vital status through 1988 had been traced through manual follow-up, 4,686 out of 5,046 (or 92.9 percent) of all deaths known to ACS volunteers were identified by the National Death Index. Since the use of automated foliow-up in CPS rr started in 1988, when there were 340 deaths from lung cancer ascertained by volunteers, another 22 have been ascertained by the use of the NDI. At a false-negative rate of 7% for the automated procedure, less thad two deaths would have been missed in our study (i.e. 0.07"22=1.54), by using the automated procedure instead of ascertaining deaths by ACS volunteers . As noted earlier, follow-up of vital status is complete for 99.8 % of all enrolled subjects, and of those 101,541 deceased subjects only 3,258 (3.2%) did not have a death certificate (ACS: Update of the CPS-II Master Index Vital Status report, April 12, 1993). ii. Results of Validation Study of Death Certificate Diagnoses of Lung Cancer '~ For 30 deaths for which lung cancer was considered the underlying cause of death in CPS II nonsmoking participants who resided in SEER registry areas, SEER diagnosis was compardd with the underlying cause of death on death certificates. In 29 subjects classified as primary lung cancer by death certificates, SEER Cancer Registries also diagnosed primary lung cancer in 27, and for two, the primary site was listed unknown in the SEER database. For no cases was the

disease known to be metastatic from other sites to the lung. I.n 25 of these 29 I lung cancers (86.2 percent), the specific histologic type was known to the SEER Cancer Registries, and in 64 percent they were adenocarcinomas. From this small validation study we conclude that lung cancers coded from death certificates generally correctly classify deaths fro~ primary lung cancer. The confirmation rate was 93. I percent (27/29), similar to that found in the TNCS study (Percy I981). Even in the two instances in our validation study in which the primary site of cancer was unknown, the diagnosis of lung cancer was not ruled out. Main Exposure Variables i. Self Reported ETS Exposure in CPS II and NI=IIS As mentioned above, because the CPS II questionnaires did not require respondents to complete all fields, many questionnaires contained blanks (Table 12). Twenty-three percent of the questionnaires filled by men and thirteen percent of those f'Llled by women were left blank in the three spaces provided for self-reported number of hours exposed to ETS (i.e., at home, work and other places). Table 13 displays in detail the patterns of answers from CPS II enrollees to the questions: " Whether or not you smoke, on the average, how many hours a day are you exposed to cigarette smoke of others? At home? (hours); At work (hours); In other areas? (hours)". As shown in this table, most times a space was left blank when valid answers were provided for at least one of the three environments. 0 0~ 0'~ 0 0'~ 0 I I I I I I I I I I I I I i I I

I I I 65 I I I I I I I I I I I I I I Table 12. Answers to question in CPS II on reported hours of ETS exposure at different settings a. All Men in CPS II Hours Home % Work % Other % Places 0 196,031 38.5 I24,114 24.4 101,533 20.0 I I5,107 3.0 52,856 10.4 71,040 14.0 2 12,258 2.4 24,834 4.9 26,533 5.2 3 8,649 1.71 9,089 1.8 6,979 1.4 4 18,363 3.6 13,359 2.6 6,240 1.2 5 8,678 1.7 5,059 1.0 1,714 0.3 6 11,904 2.3 8,527 1.7 1,384 0.3 7 2,621 0.51 3,985 0.8 280 0.1 8+ 37,827 7.4 80,478 15.8 4,710 0.9 Blank 180,924 35.6 163,357 32. i 261,885 51.5 Unclass 16,232 3.2 22,936 4.5 26,296 5.2 Total 508,594 100.(~ 508,594 100.0 508,594 100.0 All three fields left blank 65,999 (13.0%) All three fields with unclassifiabIe data 5,006 (1.0%) ~0 0 O~ 0~ 0 I

66 I I I I b. All Women in CPS II 'i:Iours Home % Work -% Other % ' Places 0 208,404 30.8 [54,373 22.8 99,953 14.8 1 16,103 2.4 40,117 5.9 46,088 6.8 2 14,029 2. I 16,636 2.5 20,148 3.0 3 10,952 1.6 6,243 0.9 7,06 [ 1.0 4 20,430 3.0 8,937 1.3 4,726 0.7 5 13,642 2.0 4,766 0.7 1,385 0.2 6 15,753 2.3 7,204 1.1 928 0.1 7 4,097 0.6 6,630 1.0 181 0.0 8+ 59,412 8.8 59,133 8.7 4,393 0.6 Blank 282,326 41.7 345,165 51.0 433,178 64.0 Unclass 31,382 4.6 27,326 4.0 58,489 8.6 Total 676,530 100.0 676,530 i00.0 676,530 100.0 . All three fields left blank 156,249 (23.1%) All three fields with unclass data 6,285 (0.9%) 0 0 0 I I I I I I i I I

! I I I- I I I I I I I I I I I I I I I Table 13. Patterns of answers given by CPS [I participants to self- assessment of number of hours exposed,, to ETS ~.i.~ Men w~m~fi -' ' Home Work Other hours hours hours , • 65,999 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 I to 1 to 1 to 1 to 1 to I to 0 5,732 1 to 8' 26,097 9 * 8,993 0 l 4,422 0 0 186 0 1 to 8 130 0 9 30 1 to 8 50,206 1 to 8 0 319 I to 8 1 to 8 7,361 I to 8 9 373 9 . 9,208 9 0 25 9 1 to 8 110 9 9 1,733 7,704 0 2,334 1 to 8 2,216 9 431 11,190 65,314 28,323 4,791 1 21,994 1 19,818 I 28,225 1 1,284 654 522 299 932 31,703 0 145 1 to 8 5,311 9 167 0 2,244 0 0 4,146 0 0 0 0 I to8 0 9 to 8 to 8 0 to 8 I to 8 to 8 9 9 9 0 9 Ito8 9 9 12.98 156,249 23.10 l. 13 8,565 1.27 5.13 24,543 3.63 1.77 22,486 3.32 0.87 5,434 0.80 0.04 173 0.03 0.03 100 0.01 0.01 80 0.01 9.87 47,135 6.97 0.06 186 0.03 1.45 4,265 0.63 0.07 745 O. I 1 1.81 9,946 1.47 0.00 18 0.00 0.02 l I9 0.02 0.34 2,282 0.34 1.51 18,908 2.79 0.46 4,372 0.65 0.44 3,743 0.55 0.08 2,613 0.39 2.20 27,148 4.01 12.84 72,162 I0.67 5.57 22,997 3.40 0.94 13,632 2.01 4.32 19,751 2.92 3.90 8,292 1.23 5.55 10,249 1.51 0.25 1,871 0.28 0.13 841 0.12 O. 10 354 0.05 0.06 265 0.04 0.18 1,206 0.18 6.23 77,326 11.43 0.03 142 0.02 1.04 6,890 1.02 0.03 876 O. 13 0.44 5,009 0.74 0.82 4,022 0.59 67 o o~ 0~ 0~ o o~ o

68 I I to 8 0 1 to 8 3,050 0.60 2,810 0.42 1 to 8 0 9 174 0.03 505 0.07 1 to 8 1 to 8 . 46,980 9.24 44,622 6.60 1 to 8 1 to 8 0 2,897 0.57 1,518 0.22 i to 8 1 to 8 1 to 8 17,191 3.38 8,415 1.24 1 to 8 1 to 8 9 791 0.16 1,316 0.19 1 to 8 9 . 206 0.04 432 0.06 1 to 8 9 0 t7 0.00 19 0.00 1 to 8 9 1 to 8 123 0.02 96 0.01 1 to 8 9 9 262 0.05 420 0.06 9 . 5,362 1.05 14,569 2.15 9 0 15 0.00 15 0.00 9 1 to 8 69 0.01 133 0.02 9 . 9 1,079 0.21 3,735 0.55 9 0 37 0.01 71 0.01 9 0 0 27 0.01 76 0.01 9 0 1 to 8 11 0.00 23 0.130 9 0 9 39 0.01 131 0.02 9 1 to 8 390 0.08 922 0.14 9 1 to 8 0 I0 0.00 II 0.00 9 1 to 8 I to 8 137 0.03 62 0.01 9 I to 8 9 211 0.04 306 0.05 9 9 3,586 0.71 4,815 0.71 9 9 0 26 0.01 28 0.00 9 9 1 to 8 227 0.04 200 0.03 9 9 9 5,006 0.98 6,285 0.93 Total 508,594 100.130 676,530 100.00 *a 9 code means that unquantifiable answ~rs(~vor~g like a ,,10t,,or "little"), as well as question marks, were answered. The comparisons of CPS-II data on ETS exposure at home, with data from the 1988 National Health Interview Survey (NCHS NHIS 1988) is presented in table 14, stratified by age, race and gender. If spaces left blank for number of hours exposed to ETS at home in CPS II are considered.to represent zero hours (i.e., unexposed), and persons with "unclassifiable" ETS information are excluded, then the prevalence figures from self reported data on ETS exposure at home in CPS rr resemble the prevalence in NHIS. Indeed, all age-adjusted comparisons of gender and racial specific prevalence figures agree within 3.3 percent. The category of "unclassifiable" ETS represent vague wording (e.g., a question mark, 'Iittle') that could not be converted into hours during coding of questionnaires. We concluded that when self-reported ETS exposure in CPS II o 0~ o 0~ o o~ I I I ,|

I I- I I I I I I I I I I I I I I 69 I was left blank, persons should be considered unexposed, and that "unclassifiable" data on ETS exposure in the three blanks should be excluded from the analyses. Table 14. Percentage of nons~aokers reportedly exposed to ETS at home* in CPS II ** and NI:IIS *** by age, race and gender. White Men Black Men White Black Women Women Age CPS NI-IIS :kDiff CPS NHISI :k CPS NHIS :k CPS NHIS _.+ lI lI Diff lI Diff II Diff 30-34 12.8 10.2 2.6 12.9 II.6 1.3 21.2 15.9 5.3 22.9 21.8 1.I 35-39 10.5 9.2 1.3 12.7 I2.4 0.3 I9.9 14.9 5.0 20.1 18.1 2.0 40-44 8.7 7.7 45-49 9.1 7.3 t.0 t4.4 9.4 1 5.0 1.8 9.6 18.1 -8.5 19.3 15.3 4.0 22.4 20.7 1.7 18.9 17.6 1.3 21.1 18.4 2.7 50-54 8.7 15.5 -6.8 I1.0 14.2 -3.2 18.3 18.6 -0.3 19.4 30.0 -10.6 55-59 8.3 14.4 -6.1 12.0 17.9 -5.9 16.4 11.5 4.9 17.1 22.8 -5.7 60-64 7.3 11.4 -4.1 10.3 25.01 -14.7 13.1 13.0 0.1 15.5 24.0 -8.5 65+ 5.2 5.9 -0.7 5.4 12.0 -6.6 8.4 7.2 1.2 11.7 11.7 0.0 Total 7.8 9.4 -1.6 9.9 13.4 -3.5 14.7 12.8 1.9 17.0 19.2 -2.2 8.8§ 9.8 -t.0 t0.9 14.2 -3.3 15.8 13.2 2.6 18.5 19.9 -t.4 ETS exposure as self-repoSed'- ~umb~ oi"h~'~'~f'~xt~osurc to ETS at home in CPS IL m~d as living with a smoking person who smokc~ at home in NHIS. ** Excludes "unelassifiable" ETS exposure at home. Considers I-8 hours as exposed, and blanks in spaces provided to write ETS exposure at home, as well as O's as unexposed. *** Weighted percentages (i.e., weights are inverse of selection probabilities) §Age adjusted prevalence figures using the 1980 US Census sub-populations as standards I

7O Comparisons of CPS II participants in the analytic cohort for self-reported ETS analyses were conducted to contrast characteristics such as age, schooling and 'race', for individuals who filled all three spaces and those who left spaces blank. Those who left any space blank were more likely to be older, and less educated, and more likely to be non-whites than those who f'ti1ed the three spaces (Table 15). However, persons who f'tlled all three fields for hours of exposure at home, work and other places and who reportedly had zero hours of exposure to ETS, were similar to those who left any blank space for ETS in CPS II questionnaires. As will be discussed in Chapter 8, a possible implication of this distribution of missing data is that perhaps blanks might not represent ETS unexposed subjects. Table 15. Characteristics of CPS II nonsmokers in analytic cohort for self-reported ETS by completeness of the information provided for ETS Characteristic Left any ETS field Completed all ETS blank fields Cross-product (Column percent) (Column percent) . rati~ Age group 65 + 30-64 Schooling <12 years 12+ 'Race' Non-whites Whites 25.4 20.5 74.6 79.5 15.6 8.1 84.4 91.9 8.3 5.5 91.7 94.5 1.4 2.1 1.7 ! !

I I I I I I I I I I I I I I I I I 71 ii. CPS II Self-reported Exposure to ETS and Spousal Smoking Habits Results of the second validationlstudy that compared self-reported ETS exposure with the smoking status of cohabitees and spouses are presented in tablel6. Table 16.a. and 16.b show that seE-reported exposure to ETS at home by CPS nonsmoking women and men, respectively, agreed with having at least one current smoker among cohabi~ees: the observed agreement was 88.4% for. women, and 94.5% of men (k=56.0%; 95% CI=55.6-56.45 for women, and k=63.5%; 95% CI=62.7-64.3 for men). Self reported ETS (hours of exposure at home) agreed better with the smoking status of spouses (Table 16.c. and 16.d.) than with the number of ~moking cohabitees (Table 16.a. and I6.b.); the observed agreement was 87.8% and 95.4% for wives (Table 16.c.) and husbands, respectively (k=62.6%; 95% CI=62.2-62.9 for nonsmoking wives, and k=69.8%; CI=69.0-70.6, for nonsmoking husbands). We concluded that self-reported ETS exposure in CPS II was internally consistent with the smoking habits reported by spouses. We also concluded that self-reported ETS is closer to spousal ETS than to smoking of cohabitees. Using current smoking status of spouses as standard, self-reported ETS would misclassify 4.6% of the subjects, with a specificity of 98%. Table 16.a. Comparison of self-reported exposure to ETS at home by CPS II nonsmoking women and the number of ........ _current smokers among cohabitees. I Cohabitees status Self-reported At least one Nonsmoker and Total ETS current smoker former smokers .................... only, Yes 33,951 (9.8) 17,250 (5.0) 51,201(14.8) No 22,850 (6.6) 271,947 (78.6) 294,797 (85.2) Total 56,801 (16.4) 289,197 (83.6) 345,998 (100.0) kL-~6~0% (¢5% CI=55.6-56.4) .......... I 0 0 ~0

Table 16.b. Comparison of self-reported exposure to ETS 72 I at home by CPS II nonsmoking men and the number of _ ....... curren,t,smokers amon cohabitees Cohabitees status Self-reported At least one Nonsmoker and Total ETS current smoker former smokers _ ...... °n!y .... Yes 6,981 (5.6) 2,814 (2.2) 9,795 (7.8) No 4,204 (3.3) I 11,622 (88.9) 115,826 (92.2) Total 11,I85 (8.9) 114,436 (91.1) 125,621 (100.0) k=63.5% (95% CI=62.7-64.3) Table 16.c. Comparison of self-reported exposure to ETS at home by CPS II nonsmoking wives and the smoking status of their husbands Husband status Self-reported Current smoker Nonsmoker Total ETS and former smoker Yes 31,945 (14.2) 5,463 (2.4) 37,408 (16.6) No 22,047 (9.8) 165,781 (73.6) 187,828 (83.4) Total 53,992 (24.0) 171,244 (76.0) 225,236 (100.0) ..... ~--£6-2~-6~/,~-(95 % ci=62.2-62.~) ..................... Table 16.d. Comparison of self-reported exposure to ETS at home by CPS II nonsmoking husbands and the smoking status of their wives Wife Self-reported Current smoker • ETS status Nonsmoker and former smoker Total Yes 6,266 (6.0) 1,741 (t.7) 8,007 (7.6) No 3,058 (2.9) 93,549 (89.4) 96,607 (92.4) Total 9,334 (8.9) k=69.8% (95% ci=69.0-70.6)" 95,290 (91.1) 104,614 (100.0) 0 O~ o~ O~ ~0 0 0"~ ~ . 0 I I I I ! I I I I I I I I I

I I I I. I I I I I I I I Among nonsmok~ug women, we also compare.d the amount of cigarettes smoked by their male cohabitees and the self-reported number of hours exposed to ETS at home. As showrt in table 17 ~d figure 6, there is a concomitant variation of more hours of exposure to ETS and the number of cigarettes and pack of cigarettes reportedly smoked by their husbands. • ~ F,g~.~re 6. Percentage of nonsmoking women exposed to spedfied self-reported number of hours of ETS at home and number of dgarettes smoked by their husbands I ~100 Amount of Nonsmoking .~ ~~/ Cigarettes 1-9 ~~0 smoked by ' °~°p~a~ck ~'7~+ H°urs Husbands , 20-3; ~2 3 4 ~ I 0 " I I I I I I Number of *Restricted to nonsmoking women whose spouses were current cigarette smoking I 0 O~ 0 O~

74 Table 17. Distribution of reported hours of exposure to ETS at home by nonsmoking women, according to number of cigarettes smoked by their husbands. Cigarettes husband smoked 0 Non smoking lto9 10 to 19 I pack 20-39 2+ packs Number of Hours of ETS at home % 1 % 2 % 3 % 4 % 75299 89.9 69l 26.3 334 13.1 135 5.91 121 2.84 2211 2.64 388 I4.8 235 9.24 139 6.09 2081 2.49 520 19.8 500 19.7 368 16.1 2484 2.97 620 23.6 792 31.2 775 33.9 905 1.08 268 10.2 397 15.6 485 21.2 745 0.89 140 5.33 284 11.2 382 16.7 83725 I00 2627 1~0 2542 100 2284 160 3.75 550 12.9 1430 33.5 1065 25 939 22 4265 100 5 % 6 % 7 % 8 % Total % Non smoking Ito9 10 to I9 I pack 20-39 56 1.79 58 1.72 13 1.52 146 2.19 76853 70.2 77 2.46 85 2.52 II 1.29 203 3.05 3509 3.21 325 10.4 291 8.62 62 7.24 626 9.4 5323 4.86 966 30.9 999 29.6 227 26.5 2094 31.4 10387 9.49 892 28.5 1015 30.1 288 33.6 1720 25.8 7035 6.43 2+ packs 809 25.9 927 27.5 255 29.8 1874 28.1 6355 5.81 Toml 3125 I~0 3375 100 856 100 6663 100 109462 100 I I ! I ! I ! I

I I I 75 I Chapter 6: Descriptive Statistics Variables Frequency of Self-reported and Spousal ETS of Exposure Forty-eight percent of the nonsmoking population in our analytic cohort reported ETS exposure at home, work: or other places. Table 18 presents the distribution of self-reported number of hours of ETS exposure at home, work: and other places, and combined in the three settings, according to the definitions presented in Section 3.9. Fourteen percent reported any exposure at home, 26 percent at work: and 18 percent from elsewhe~. Among those exposed to any ETS, one third was exposed to ETS for one or two hours, another third was exposed for two to five hours, and the rest to six and more hours of ETS. Accordingly, cutoffs of ETS were used at 3, and 5 hours of self-reported exposure at home, I, 2 and 6 hours of self-reported exposure at work:, and 1, 2, and 3 for self- reported exposure elsewhere, to create categorical variables and conduct further analyses. Up to 9.7 percent of nonsmokers had 3 and more hours of ETS exposure at home, but only 2.6 percent obtained that amount of exposure to ETS in places other than work or home. I More than half of the nonsmoking spouses, or 53.6 percent, in the analytic cohort for spousal ETS were married to smoking spouses. As mentioned before, smoking of tobacco products other than cigarettes were not collected in the questionnaires sent to wome[~, and thus all the spousal smoking of nonsmoking husbands comprised exclusively cigarette smoking. On the other hand, 33.7 percent of nonsmoking wives (or 71, 891) were married to nonsmokers, and two-thirds, or 66.3 percent, were married to ever smoking husbands. The latter group could b~ further divided according to the following types of smoking: 15.8 percent ( or 33,705) were married to current cigarette smokers; 30.1 percent (or 64,230) to former cigarette smokers; 2.5 percent (or 5, 487) to smokers of both cigarettes and pipes or cigars; 4.6 percent (or 9,794) to current pipe and or cigar smokers [vho formerly smoked cigarettes; 6.7 percent I

76 or (14,306) to former smokers of both cigarettes and pipe/cigars; 4.3 percent (or 9,253) to former pipe/cigar smokers who never smoked cigarettes; and 2.1 percent (or 4,487) to former cigarette smokers who then smoked pipe or cigars. Correlates of ETS exposure However, the contributions of each ETS exposure setting to the overall exposure varied greatly by gender. More men more than women reported exposure at work, whereas women reported most of their exposure at home. Table 18. Hours of exposure to ETS reported by nonsmoking CPS II participants at different settings, 1982 a. Both Men and Women Places Total ETS None 337,144 86.0 288,832 73.6 321,012 81.8 205,433 52.4 1 9,855 2.5 37,737 9.6 46,554 11.9 61,490 t 5.7 2 7,324 1.9 14,039 3.6 14,379 3.7 31,140 7.9 3 8,679 1.4 4,911 1.3 4,202 1.1 13,116 3.3 4 5,367 1.4 6,376 t.6 2,512 0.6 t3,366 3.4 5 5,748 2.2 2,914 0.7 745 0.2 8,074 2.1 6 5,748 1.5 4,355 I. 1 433 0.1 9,051 2.3 7 1,290 0.3 3,349 0.9 102 0.0 5,010 1.3 8+ 11.484 2.9 29.713 7.6 2,287 0.6 45.546 11.6 Total 392,226 I00 392,226 1t30 392,226 1130 392,226 o 0~ o 0~ ! 1 I i I ! ! I I I I I I I

I I I 77 I I I I I I I I I I I I I I I b. Men Hours Home % .... Work % Other % Total % Places ETS None 101,481 91.7 69,9~3 63.2 81,912 74.0 48,175 43.5 1 2,641 2.4 l 6,909 15.3 20,758 18.8 24,272 21.9 2 1,488 1.3 6,363 5.7 4,990 4.5 12,548 11.3 3 933 0.8 2,072 1.9 I, 144 1.0 4,355 3.9 4 1,380 1.2 2,7~4 2.5 783 0.7 3,937 3.6 5 606 0.5 989 0.9 289 0.3 2,009 1.8 6 726 0.7 1,513 1.4 135 . 0.i 2,192 2.0 7 93 0. I 704 0.6 39 0.0 1,042 0.9 8,~- ~ l,.340 1.2 9.486 8.6. ..638 0.6 12,158 !I.0 Total 110,688 I00 110.688 I00 110,688 100 110,688 I00 e. Women Hou~ ....... H~in~ % Work % Other % Total ETS % : I Places None 235,663 83.7 218,859 77.7 239,100 84.9" 157,258 55.9 1 7,214 2.6 20,828 7.4 25,796 9.2 37,218 t3.2 2 5,836 2. I 7,703 2.7 9,389 3.3 l 8,592 6.6 3 4,402 1.6 2,83~ 1.0 3,058 1.1 8,761 6.6 4 7,299 2.6 3,652 1.3 1,729 0.6 9,429 3.3 5 4,761 1.7 t,925 0.7 456 0.2 6,065 2.2 6 5,022 1.8 2,842 1.0 298 0.1 6,859 2.4 7 I, 197 0.4 2,645t 0.9 63 0.0 3,968 1.4 8+ 10.144 3.6 20,245 7.2 1.649 0.6 33,388 11.9 Total 281.538 I00 281,538 1t30 281.538 L00 281,538 lt30 I As shown in table 19, self-reported exposure to ETS decreased from 70 percent to 16 percent with increasing age. Spousal smoking exposure, however, did not show this trend, except for current smokers (Table 20). One implication of this ! o P0 o

78 difference, as will be discussed in Chapter 8, may be that self-reported ETS does not reflect long-term, but rather current ETS exposure. Table 19. Proportion of nonsmoking CPS II participants in analytic cohorts who reported any ETS exposure by age at .......... interview,h. 1982 ..................... ...... Age Group . Men and Women .Women ................. Men_ .. 30-34 70.1 67.5 75.1 35-39 66.5 64.5 71.9 40-44 61.0 58.9 70.6 45-49 61.6 57.5 70.7 50-54 58.6 54.7 67.5 55-59 52.9 49.1 62.3 60-64 43.3 39.5 52.6 65-69 29.6 26.7 36.6 70-74 20.7 18.7 26.2 75+ years 15.5 14.7 18.1 Total 52.4 55.9 43.5 Table 20 presents the proportion of participants with spouses also in the study and whose spouses were ever smokers (i.e., spousal smoking). Almost two- thirds of nonsmoking women lived with ever smoker husbands whereas only 27 percent of men were married to ever smoker wives. Table 20..Proportion of nonsmoking participants married to ............ eve_r-smokin.g.....sP0use by age at, interview, 1982 Age Group Men and Women Men Women 30-34 46.5 53.9 27.2 35-39 52.7 60.7 29.2 40-44 57.4 62.7 31.3 45--49 52.8 63.7 29.8 50-54 54.8 67. i 29.0 55-59 56.7 69.4 28.1 60-64 54.3 68.3 26.1 65-69 51.2 67.4 23.4 70-74 49.0 66.6 21.4 ~75+ ~¢ears . 40.5 63.0 16.1 Total -53~6 .............66.3 " " 26.8 I I I I I I I I I I I

| I I I i I I I I I I I I I I I I 79 Important correlates of ETS exposure were schooling and race, both of which can be taken as surrogates of social class. Non-white men and women were slightly less self-reportedly exposed than whites: the ratio of the proportion of I exposed nonwhites to exposed whites was 0.8. Married women were 5.8 percent more likely to be self-reportedly exposed to ETS than unmarried women. More educated men and women were more likely to report ETS exposure, or to have an ever-smoker spouse than less educated men and women: the ratio of self-reported ETg exposure among the most formally educated, to that among the least formally educated was 2.8 and 2.0 for men and women, respectively. However, among husbands and wives, the ratio of smoking spouse among the most educated the least educated was 0.6 for men and women. Unlike self-reported ETS exposure, assessment of ETS exposure based on the smoking history provided by spouses in the study shows a picture consistent with the demography of ETS exposure in the US population at large that was described in Chapter 1. Therefore, the assessment of ETS exposure based upon 1 spousal ETS might reflect better true ETS exposure, than self-reported ETS. People reporting exposure to asbestos, chemicals, coal dust or tar, formaldehyde and ionizing radiation, were more likely to report ETS exposure but did not / differ substantially according to the smoking status of spouses. ETS exposed and unexposed were comparable with respect to medical history of any chronic non-malignant disease. ETS exposed and unexposed groups were also comparable with respect to the consumption of foods considered major sources I of carotenoids. 0~

Table Characteristic Number of subjects Age at interview (years) (standard deviation) Rac~ (% white) Married (%) Education (%) <High School High School Trade School or some College L>College Occupation (%) Any asbestos Other lung carcinogens§ Diet: times/week (standard d~viation) G~e~n leafy vegetables FruitHuices Characteristics of nonsmoking participants accordiag to self-reported ETS Men Women No ETS An,v ETS No ETS Any ETS 48,175 62,513 157,258 124,280 (43%) (57%) (56%) (44%) 60.4 53.6 60.2 53.2 (11.2) (9.6) (I 1.7) (10.0) 92.9 94.3 92.1 93.3 93.7 93.4 74.1 79.9 17.7 7.6 18.3 9.9 19.9 17.7 31.1 33.8 21.7 24.8 2&7 30.5 40.7 49.8 23.9 25.8 4.9 7.1 1.4 2.4 20. l 26.6 6.8 11.9 4.8 4.7 5.0 5.0 (2.0) (2.0) (2.0) (1.9) 5.2 5.0 5.5 5.4 (2.3) (2.3) (2.1) (2.2) Chronic lung dis. (%) Any 7.2 7.2 7.6 8.0 Tuberculosis I. I 0.9 1.0 1.0 Chronic bronchitis 1.7 1.7 2.9 3.1 Emphysema 1.0 0.5 0.5 0.3 Asthma 4.4 4.9 4.2 4.6 § Self-reported occupational exposures to: Chemicals, coal dust or tar, formaldehyde and ioni~.ing radiation. 80 o 0", r~ o o~ I I I I I

I I I I- I I I I I I I I I I I I I I Table 22. C'hamct~dstic Number of subjects Age at interview (years) (standard deviation) Race (% white) Mar~ed (~,) Education (%) <High School High School Trade School or some College _:>College Occupation (%) Any asbestos Other lung carcinogens§ Diet: dmes/w~k (standard deviation) Green leaf3, vegetables Fruit/Juices CharacteriStics of nonsmoking participants according to spousal smoking Husbands Non smoker , I 73,914 ~7.4 (10.2) 94.9 I I00.0 Wives W'fie Non smoker Husband ever husband ever smoked smoked 27,040 71,892 141,262 55.8 54.5 55.3 (9.4) (10.1) (9.5) 94.7 94.8 95.0 ~oo.o ~oo.o [oo.o 12.6 8.5 9.0 I 1.5 20.2 ; 16.2 31.3 36.0 23.9 23.2 30.1 29.6 43.2 52. I 29.5 23.0 6.2 6.6 1.7 1.8 24.3 24.2 8.9 9.4 4.0 (2.5) (2.3) 4.0 4.7 4.6 (2.5) (2.2) (2.3) 5.2 5.4 5.4 (2.3) (2.1) (2.2) Chronic lung dis. (%) Any 7.1 7. I 7.2 7.7 Tuberculosis 1.0 [ 1.0 0.9 1.0 Chronic bronchitis 1.7 1.4 2.6 3.0 Emphysema 0.8 0.6 0.3 0.3 Asthma 4.5 4.9 4.3 4.4 ~ Sel~S~eportcdr~cupati0nal ex~posure~' ~0: Chemicals, coal dust or t~r, formaldehyde and ionizing radihtion. 81

82 i Smoking status, quantity and duration of ETS from spouses Spouses of subjects in the study comprised the following major categories: I) nonsmoking spouses (145,806, or 46.4 percent) 2) current and former cigarette smokers with .complete data (41,099 or 13.1 percent, and 71,594 or 22.8 percent respectively), 3) former and current smokers with incomplete data (1,764 or 0.6 percent and 6,087 or 1.9 percent, respectively), 4) ever smokers with unclassifiable smoking (4,431 or 1.4 percent) and 5) ever pipe and or cigar smokers which includes a mixture of smoking of the different tobacco produ6ts (43,327 or 13.8 percent). Analysis of dose-response relationships between ETS from spousal smoking and the risk of cancer among nonsmokers is restricted to cigarette smoking spouses with complete data, and the univariate statistics of the variables used in the analyses are presented. The quantity of smoking is based upon frequency (i.e., cigarettes per day) as recalled by the smoking spouses of nonsmoking participants. We set a value of zero for nonsmoking spouses of nonsmokers in the study. The number of cigarettes smoked by the 41,099 spouses of nonsmoking participants who were current cigarette smokers and had complete smoking information, and who never smoked cigars or pipes, ranged from I to 100 a day, with a mean of 22.6 cigarettes per day, a first quartile of 15 cigarettes, a median of 20, and a third quartile of 30 cigarettes. Ninety percent of current smoking spouses smoked up to two packs of cigarettes. The position of the fast and second tertiles for the quantity of cigarette smoking among the current cigarette smokers of nonsmokers was 20 (i.e., one pack) and 25 cigarettes per day, respectively. Among the 71,594 nonsmokers in this analytic cohort whose spouses were former cigarette smokers, had complete smoking information, and did not smoke other tobacco products, the quantity of usual former cigarette smoking were O~ O~ O~ 0 I 1 1 1 ! 1 I I I I I I i I I

I I I I I I I I I I I I I I I I I I 83 similar to those previously shown for current cigarette smoking spouses: a range from I to 100 cigarettes per day, the mean was in 22.6 cigarettes per day, the median was 20 (i.e. one pack), but the first quartile was lower at I 1 cigarettes per day, and the third quartile was 30. Again, ninety percent smoked less up to two packs of cigarettes. The tertiles were 18 and 20 for the lower and upper, respectively. For those subjects in the analytic cohort of spousal ETS who were married to cigarette smoking spouses with complete data, we computed the time they were married to spouses, to assess the effect of this variable as well as to compute the pack-years smoked during marriage by smoking spouses. As described before, we excluded those subjects married more than once or whose spouse was also married more than once, or with incomplete data on age at marriage., since both were needed to compute time in marriage exposed to smoking spouse (i.e., the difference between age at interview and the age at first marriage yielded the duration of marriage, and is used in combination with the age at uptake and quitting smoking as well as the age at interview of smoking spouses, to compute the time during marriage nonsmokers were exposed to spousal smoking). Typically nonsmoking husbands and wives who were married to a cigarette smoking spouses, had spent in average 21 years (standard deviation of 12 years, median=21 years) exposed to ETS, and the values of this variable ranged from 1-63 years, and differences between men and women were small. The cutoffs for the tertiles of the distribution of the duration of exposure to ETS from spousal smoking for these group of individuals were 15 and 27 years for exposed men, and 17 and 30 for exposed women. The distribution of study subjects according to the combination of usual amount of cigarette smoking with the duration of exposure to ETS from spousal cigarette smoking during marriage is presented in Table 23. Among nonsmokers married to smoking spouses pack-years ranged from I to 198 with a mean of 24.0 pack-years, and a median of 20 pack-years, and the cutoffs of tertiles the cutoffs were approximately 16, and 35, but men and women differed considerably: the mean

84 packyears of E-tS spousal smoking among nonsmoking men and exposed to any Ei-S were 19 and 29, respectively. A t-test yielded a p-value <0.000I. Table 23. Study populations included in spousal analyses for intensity and duration of ETS exposure from spousal cigarette smoking among nonsmoking spouses in CPS II a. Smoking of current and former cigarette smoking (i.e., nonsmokers =0) n=258,499, see page 88. ..... ~0unt of "' Amount of current Number % former Number % .... s_mokin~g ............ ~moking 0 145,806 78.0 0 145,806 67.1 < I pack 12,606 6.7 < I pack 23,917 11.0 1 - t.9 packs 21,511 11.6 i - 1.9 packs 33,878 15.6 ..... 2_+. packs 6,982 3.7 .2+ packs 13,799 6.3 Total 186,905 100.0' Total 217,400 i00.0 Results in table37. " " ~-b. Time in marriage With cigarette smoking spouse, and packs of cigarette-years in marriage, (nonsmoking spouses set to 0), n=148,204. Results in table 39. Exposed to spousal ETS for: Years Men % Years Women % Number Number None 46~0~9 852 ~)2[ None 46,149 50.0 1-15 3,326 1-17 14,794 16.0 16-26 3,125 516 ! 18-29 15,491 16.8 27+ 3,492 6.2[ 30+ 15,788 17.1 .................... :Tot~ 148,204 100.0 c. Pa~ks of-ci~-t~tt~:y~ars-in ~a'r'riage; (nonsmoking spouses set to 0), n=148,204. Results in table 40. ...... Pack- " ' Mer~ ...... % " Pack-years ..... years Num~r ..... 46,039 82.2 0 0 1-8 9-22 23+ Women % Number 76,771 83~2 3,339 6.0 1-I6 15,451 16.7 3,263 5.8 17-35 15,569 16.9 3,341 6.0 36+ 15,053 16.3 To~l 148,204 100.0 I I I I I I I I I I i I

I I I I I I I I I I I I I I I I I 85 Chapter 7: Main Results 7.1 Deaths from Lung Cance.r ~tnd Histological Data in Death | Certificates There were 362 deaths from lung cancer among nonsmokers in the study cohort of self-reported ETS exposure. Before the ND[ update of follow-up status was completed to include new deaths frorh lung cancer that occurred between October I, I988, and December 31, 1989, we reviewed the death certificates from 284 or 78 percent of the deaths from lung cancer f'mally included in our analyses. These 284 deaths were the deaths ascertained during the f'trst six years of follow- up. In 169 instances or 59 percent, d~ath certificates only mentioned lung cancer without any reference to histological type. In the remaining 115 lung cancer deaths, or 41 percent, the certificate specifically mentioned histological type. Table 24, shows the frequencies of each major histologic type. Seventy percent of lung cancer among nonsmokdrs, when their histological types were documented in death certificates, were adenocarcinomas. If the unclassified were excluded, the proportion of adenocarcinomas would be 75 percent. There were no differences by gender in the distribution of histological types mentioned in death certificates. I Table 24. Distribution of hystological types in 115 deaths from lung cancer among nonsmokers in the analytic cohorts in CPS II, 1982-1988, for which this information was readily ....... _available from ~eath .certificates • ..T._)~pe_ Women. ,Men ....... Total _ Number (%) Number (%) Number (%) Adenocarcinomas 59 (71.1) 21 (65.6) 80 (69.6) Squamous cell 12 (14.4) I 7 (21.9) 19 (16.5) Large cell i (1.2) 3 (9.4) 4 (3.5) Other types 3 (3.6) 1 (3. i) 4 (3.5) Unclassified 8 (9.6) 0 (0.0) 8 (7.0) Total 83 (100.0) I 32 (100.0) 115 (100.0) l

86 I 7.2 Potential confounders: age, gender, schooling, race, prexisting lung disease, occupational exposure to lung carcinogens, consumption of foods containing carotenoids, and fat as nutrient index. Age The rates of lung cancer increased monotonically with age. Figure 7 depicts this feature of the risk of lung cancer by age: observed values were fitted empirically by Poisson and exponential regression models, both providing an adequate description of the data. As we described earlier, age was also strongly associated with self-reported and ETS exposure from spousal smoking. Thus, age was included in all models. Figure 7. Death Rates of Lung Cancer among CPS II Nonsmoking participants by Age, 1982-1989 8O 70 60 50 Rate per 40 lOO,OOO 30 20 10 0 <50 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+ Age groups Observed ~'~ Exponential ~ Poisson o o 0~ I I I I I I I I I I / I I

87 I I I I I I I I I I I I I I Gender Nonsmoking men showed an increased risk of lung cancer as compared with nonsmoking women, the crude rate ratio being 1.2 (95% CI=I.0-1.5) (i.e., 116/ /786,532 , Men and women were comparable with respect to age, men being slightly younger (Table 25). Thus, the Mantel-Haenszel RR for gender, adjusted for age was 1.3 (95% CI= 1.1-1.7). Men had a slightly greater likelihood of being reportedly exposed to ETS than women (56 percent and 53 percent, respectively). Sources of exposures to ETS were different for men than for women (i.e., most for men from work, and women at home). Women were more exposed to ETS from, their husbands, than were nonsmoking men from their wives. Thus, gender was included in all the models. Table 25. Death rates of Lung Cancer among Nonsmokers by Age and G{mder, CPS II, 1982-1989 Men Women Age-group <50 3 121,1781 I5 2 4 50-54 4 133,136 17 3 18 55-59 4 144,161 18 3 21 60-64 16 131,477 17 12 30 65-69 27 109,5881 14 25 30 70-74 26 76,221 10 34 39 75-79 14 42,807 5 33 44 80-84 17 19,322 2 88 35 85+ 5 8,643 ~ 1 58 25 Person- % Rate t~ Person- % Rate >,e~ ...... lOT,~ ..... 7e~ ,,, _1o-5 ~ 341,834 17 1 303,204 15 6 338,198 17 6 329,075 16 9 270,185 13 I1 196,522 I0 20 129,427 6 34 69,869 3 50 41,768 2 60 Total 116 786,532 I00 15" 246 2,020,081 (n=392,226 subjects) *Age-adjusted to the CPS-rr population. Age-SRR=I.4 (95% CI=I.I-1.8) RR (m-h)=t.3 (95% CI=1.I-~1.7) I00 10"

88 Race I I I I I I ! ! ! ! ,, ! Only 9.4 percent of the deaths from lung cancer, or 34 occurred among 'nonwhites'. All but three of these deaths occurred among African-Americans. We collapsed all 'races' different from 'whites' into a category of 'nonwhites' for the purpose of the analyses. Table 26 presents deaths rates by this dichotomous variable, and by age. The total row presents age-adjusted rates. Nonwhites had a 44% increased risk of lung cancer, after adjusting for age [RR re.h=1.4 (95% CI=I.0-2.0)] (Table 26). As pointed out before, nonwhites were less likely to report ETS exposure but were comparable in the proportion of spousal ETS. Therefore, we included 'race' in the multivariate analysis of ETS and lung cancer. Table 26. Death rates of Lung Cancer among Nonsmokers by Age ,and 'Race', CPS II, 1982-1989 Whites Nonwhites Age-group ~ Person- % Rate t~hs Person- % Rate years 10-5 years 10-5 <50 0 43,513 22 0 7 419,493 16 2 50-54 2 30,306 15 7 20 406,037 16 5 55-59 2 31,258 16 6 23 451,103 17 5 60-64 1 28,440 14 4 45 432,112 17 10 65-69 10 23,123 12 43 47 356,650 14 13 70-74 6 17,487 9 34 59 255,256 I0 23 75-79 8 11,458 6 70 48 160,776 6 30 80-84 2 6,306 3 32 50 82,885 3 60 85+ 3 4,271 2 70 27 46,140 2 59 Total 34 196,161 I00 17" 326 2,610,452 100 11" (n=390,833 subjects and 360 deaths) *Age-adjusted to the CPS-II population. Age-SRR=I.6 (95% CI=1.1-2.3) RR (m-h) age-adjusted=l.4 (95% CI=I.0-2.I) 0

89 I I I I I I I I I I I I I Schooling Years of education was both related to ETS exposure in the entire study population and was a risk factor for lung cancer. Younger CPS II nonsmoking participants tended to be more educated than older, as shown in table 27, along with the corresponding number of deaths, person-years under observation, and death rates. The unadjusted comparison of rates of those who did not graduate from high school, as compared to those who did, was 2.2 (95% CI=1.8-2.8). After adjustment by age, the Mantel-Haenszel estimate dropped to 1.2 (95% CI=0.9-1.5). Years of education was positively associated with self-reported ETS, and inversely related with spousal smoking status. Although schooling was not a meaningful confounder based upon the data at hand, it was included in all multivariate analyses based upon a priori knowledge of the association between lung cancer and low socioeconomic status. Table 27. Death rates of Lung Cancer among Nonsmokers by Age and S.ehooling, .CPS II, ....1982-.!989 ..... <12 },rs 12 + ,vrs Age-~oup Dent Person- % Rate Dear Person- % Rate hs years 10-5 hs years 10-5 <50 0 20,515 6 0 7 442,497 18 2 50-54 1 27,204 7 4 21 409,136 17 5 55-59 2 41,590 I 1 5 23 440,769 18 5 60-64 4 52,470 14 8 42 408,081 17 10 65-69 13 60,234 I6 22 44 319,539 13 14 70-74 20 61,908 17 32 45 210,835 9 21 75-79 18 50,765 I4 35 40 121,468 5 33 80-84 21 31,749 9 66 31 57,442 2 54 85+ 13 23,322 6 56 17 27,089 1 63 Total 92 369,758 100 13" 270 2,436,856 100 12" (n=392,226 subjects) *Age-adjusted to the CPS-II population. Age-SRR=I. 1 (95% CI=0.8-1.4) RR (m-h) age-adj.=l.2 (95% CI=0.9-I.5) I 0 ".4

90 I Asbestos Self-report of being ever occupationally exposed to asbestos was associated with a two-fold higher risk of lung cancer among'nonsmokers in CPS lr (Table 28) [age-adjusted RR (m-h)=2.0 (95% CI=1.I-3.5)]. The effect estimate associated with asbestos was similar [multivariate RR=I.8 (95% CI=1.I-3.2)], after controlling for age, gender and the indicator of 'race', and schooling, as well as total intake of foods containing carotenoids, history of chronic lung disease and self-reported ETS exposure in Cox multivariate analyses (Table 30). Adjusted rate ratios associated with ever being exposed to asbestos at work were slightly lower for men [1.5 (95% CI=0.7-3.I)] than for women [2.3 (95% CI=I.0-5.3)]. Table 28. Death rates of Lung Cancer among Nonsmokers by Age and self-reported Occupational Exposure to Asbestos, CPS II,1982-1989 Age-group Ever exposed Unexposed Person- % Rate t~, Person- % Rate years 10-5 years 10-5 <50 0 13,913 20 0 7 449,099 16 2 50-54 2 12,326 18 16 20 424,018 15 5 55-59 0 13,073 19 0 25 469,288 17 5 60-64 2 11,730 17 i7 44 448,822 16 10 65-69 5 9,071 13 55 52 370,701 14 14 70-74 1 5,384 8 19 64 267,359 10 24 75-79 0 2,609 4 0 58 i69,625 6 34 80-84 2 999 1 200 50 88,192 3 57 85+ 2 457 I 437 28 49,953 2 56 Total 14 69,562 100 25* 348 2,737,057 100 11" (n=392,226 subjects) *Age-adjusted to the CPS-17 population. Age-SRR=2.2 (95% CI=1.2-3.9) RR (m-h) age-adjusted=2.0 (95% CI= 1. I-3.5) 0 0"~ CO 0", 0 O~ I I ! I I I I I I ! I I I I

! ! ! 91 Only 2.5 percent, or 9,664 subjects reported ever having occupational exposure to asbestos. Five percent of men in the analytic cohort for self-reported ETS fell in that category, and only 1.4 ~percent of women said they had ever been occupationally exposed in their life-time. Among those who said they had ever been exposed to asbestos at work, 65 percent were in the labor force by 1982. The occupations/industries more frequently mentioned by them included professions with presumably 10w exposure such as teaching (21 percent), management (i 1 percent), engineering (5 percent), and technicians (2 percent), as well as trades with potentially higher exposures such as automechanies (5 percent), and construction (2 percent). Twenty-five percent of those ever exposed to asbestos had retired from the same occupations. Sixty percent of CPS II participants had previously held a job different from that currently held or at retirement. The major frequencies were approximately the same (teachers, managers, and salesmen, 10, 4 and 8 percent, respectively). In addition, 2 percent mentioned they also had worked at factories, 4 percent were farmers and fishermen, and 8 percent worked in offices. Of those nonsmokers ever engaged in occupations known to carry likely high exposure to asbestos (i.e., shipbuilding, pipefitters), or likely low exposure to asbestos (i.e., plumber, construction, duckworker, autorepair, and electrician), or engaged in any occupation which fell in the category of 'possibly exposed' (See the detailed list in the Covat~iates section in 3.6), only those subjects who were reportedly ever engaged in shipbuilding trades (n=95) had a significant increased risk [multivariate RR=9.7 (95% CI=1.3-71.3)]. The number of years men and vomen were ever exposed to asbestos were grouped by tertiles of those ever exposed (i.e., the categories being 1-5 years, 6- 15 years, and 16+ years of exposure) and compared with those who reportedly were never exposed to asbestos at work. Table 29 shows the deaths, person- years and rate ratios for men an~ women separately and combined in the last column. Among the 9,664 exposed CPS II participants, 8,316 (or 86 percent)

92 I I I I I | ! ! ! reported the number of years exposed; out of the 14 deaths among nonsmokers ever exposed to asbestos, only 12 had information on years of exposure. Multivariate rate ratios showed in the third, six and nine columns of table 29, indicate that the rate ratios increase with longer exposure; however, did not follow a consistent increasing trend. The risk of lung cancer among nonsmokers exposed to asbestos in this study increased to 3. i for those who worked up to five years.as compared to nonsmokers unexposed to asbestos and then decreased to 1.2 for those who worked 5 to 16 years and remained at 1.2 among those who worked more than twenty years. Among women there was a non-statistically significant increasing trend of lung cancer risk by years of self- reported exposure to asbestos (p---0.15), but not for men (p=0.66). Self-reported occupational exposure to asbestos was a potential confounder of the ETS and lung cancer association, so we included asbestos in multivariate analysis. Preexisting chronic lung disease Medical history of any obstructive pulmonary disease (asthma, chronic bronchitis, or emphysema) or tuberculosis, or a combination of all of these conditions was not associated with the risk of lung cancer for men and women combined (RR=I.0 (95% CI=0.7-1.5)]. However, among men there was a statistically significant increased risk (RR=2.1, 95%CI=1.3-3.6), whereas among women, there was no association (RR=0.6, 95%CI=0.4-1.2). The interaction term of gender and history of chronic lung disease, when adjusting for all other covariates was statistically significant (LRz24e=t =9.2, p---0.002). This apparent effect of reported medical history of preexisting lung disease on the risk of lung cancer among nonsmokers in CPS II was heavily influenced by the history of chronic bronchitis. There was an increased risk of lung cancer among nonsmoking men with a history of chronic bronchitis [multivariate RR=3.8 (95% CI=1.8-7.8)], and there was none among women [0.6 (95% CI--0.2-1.4)]. 0 O~ Co O~ 0 O~ 0 I I

Table 29. Lung cancer rate ratios among nonsmokers by years of occupational exposure to asbes!.os, and gender, CPS II,1982-1989 Deaths Years of Deaths Men- RR* Deaths Wome RR* Men All person- RR** exposure among Years (95% among n- (95% CI) and years (95% CI) Men CI) Women Years Women None 108 745,894 1.0 240 1,9910 356 2,737,057 1.0 163 28,925 Ever 8 40,638 1.5 6 2.3 14 69,562 ! .8 (0.7-3. l ) ( 1.0-5.3) ( 1. 1-3.2) 1-5 yrs 6 12,793 3.6 1 7,456 1.9 7 20,249 3.1 _ (1.5-8.4) -. (0,3- - (1.-4-6.8) 13.4) 6-15 yrs 1 9,708 1.1 ! 9,280 1.3 2 18,988 1.2 (0.2-7.8) (0.2-9.5) (0.3-4.8) 16+ yrs 1 13,158 0.6 2 6,890 2.8 3 20,048 1.2 (0.1-4.2) (0.7- (0.4-3.9) • ,,,., 11.2) p for trend 0.66 0.15 0.23 Missing 0 4,979 2 5,298 2 10,,27.7 (n=392,058; 168 were exposed but had vague data on number of years e~Posed) *Adjusted for age, schooling, 'race', consumption of foods containing earotenoids, total fat as a nutrient index, history of chronic lung disease, and ETS. ** In addition adjusted for gender. ~890~9890~

94 Diet: foods containing carotenoids and total fat After controlling for other covariates, a reduction in the risk: of lung cancer was observed among nonsmokers with the highest reported frequency of consumption of a combination of the following foods containing carotenoids: carrots, squash and corn, green leafy vegetables, cabbage, broccoli and Brussels sprouts, tomato, and fruits andjnices. There was a borderline significant trend of decreasing risk of death from lung cancer by increasing frequency of wee~y intake of these food items (LRZ2d.f=L=3.043, p value for trend=0.0811). Nonsmokers who were in the upper quintile of the distribution of total fat intake (as nutrient index) had a statistically significant increased risk of lung cancer as compared to those in the lowest category, after adjusting for all other covariates (age, 'race', gender, schooling, history of lung disease, frequency of consumption of foods containing carotenoids, and occupational exposure to asbestos). There was a statistically significant dose-response relationship between the risk of lung cancer among nonsmokers by increasing level of total fat intake (LRZ2df=L=4.695, p value for trend=0.0302). There was a weak positive correlation between the frequency of consumption of foods containing carotenoids and total dietary fat intake (rxy=0.29). The partial correlation coefficient controlling for schooling, age, gender, was essentially unchanged (rxylZ 1 ,z2,..Zp=0.30). Other risk factors Nonsmokers who had ever been occupationally exposed to ionizing radiation showed a non-statistically significant increased risk of lung cancer [multivariate RR=I.6 (95% CI=0.7-3.5)] (Table 30). No evidence was found of an increasing trend of years of self-reported occupational exposure to ionizing radiation, when those ever exposed where grouped by tertiles of years of exposure to ionizing radiation at work: the multivariate RR were 0.9 (95% CI=0.2-3.7), 1.1(95% CI--0.3--4.5), and 0.9 (95% CI=0.2-3.6), for the first, second and third tertiles, respectively. The p value for a test for trend was 0.9. 0 0 I I I I I I

95 I I I I I I I I I I I I I I Other occupational exposures to lung carcinogens such as formaldehyde, coal tar products, and chemicals, as recorded in CPS II questionnaire, were not associated with the risk of lung cancer among nonsmokers [multivariate RR=0.9 (95% CI---0.6-1.3)] (Table 30). Multivariate rate ratios on the covariates presented in this section arc summarized in table 30, and are all included in the model, along with self-reported ETS : age at interview grouped into nine five-year groups, gender, schooling and 'race' and history of chronic lung disease as dichotomous variables, frequency of consumption of foods containing carotenoids grouped into tertiles compared with no consumption of foods and vegetable, and total fat as a nutrient index grouped into quintiles. In addition, indicator variables were included for missing observations on 'race', schooling, and diet. As shown in table 30, multivariate Cox regression analysis shows that when smokers under 50 years of age were used as referent, the RR estimates increased monotonically by every five-year age period: .1.8, 2.9, 4.4, 6.1, 8.2, 14.6, 14.6, and 22.4 . Men had a 30% increased risk as compared to women [RR=I.3 (95% CI=1.0-1.6)]. Non-whites had a 50% increased risk of lung cancer as compared to whites [RR=I.5 (95% CI=1.1-2.2)]. Asbestos was associated with almost a two-fold increased risk ['RR=I.8 (95% CI=1.1-3.2)]. History of chronic lung disease was not associated with the risk of lung cancer among men and women together [RR-1.0 (95% CI= 0.7-1.5)]. Consumption of six groups of vegetables and fruits/juices was associated with a 30% decreased risk of lung cancer but showed no clear pattern of dose-response relationship. Subjects classified in the upper 20% of the distribution of intake of total fat as a nutrient index had a 70% increased risk of lung cancer;, the intake of fat showed a statistically significant increasing trend with lung cancer death risk. 0 0

96 Table 30. Risk ratios of lung cancer among nonsmoking CPS II participants by nonTET,,S.,kno,~n risk fa~t0rs' I I I I i | ! Multivariate § Risk Factors RR (95% C.I.) Aie g 6up <50 1.0 50-54 1.8 (I.0-3.0) 55-59 2.9 (1.8-4.7) 60-64 4.4 (2.7-7.0) 65-69 6.1 (3.8-9.9) 70-74 8.2 (5.0-13.3) 75-79 14.4 (8.8-23.9) 80-84 14.6 (8.1-26.6) 85+ 22.4 (I 1.6-43.6) Schooling 12 yrs. + 1.0 <12 yrs. 1.2 (0.9-1.5) Asbestos at work i .0 Never 1.8 (1.1-3.2) Ever Ionizing radiation at work Never 1.0 Ever 1.5 (0.7-3.5) Risk Factors Other occupational exposures to lung carcinogens* None Any Gender Multivariate § RR (95% C.I.) Women 1.0 Men 1.3 ( 1.0-1.6) Whites 1.0 Non-whites 1.5 ( 1. I-2.2) Frequency of consumption of carotenoid containing foods None 1.0 Seldom to 2/week 0.3 (0.1-0.8) 3 week 0.3 (0.1-0.7) >3 week 0.3 (0.1-0.7) p for trend 0.08 Total fat as nutrient intake in quintiles Least 1.0 2 1.2 (0.8-1.6) 3 1.3(0.9-1.8) 4 0.9 (0.6-1.3) Most 1.7 (1.2-2.3) p for trend 0.03 History of chronic lung disease None 1.0 Any of these: 1.0 (0.7-1.5) Tuberculosis 1. I (0.4-2.6) Emphysema 1.8 (0.8-4.2) Asthma 1.7 (0.9-3.6) Chronic bronchitis 1.2 (0.7-2.1) (n=392,226) § Cox regression models included age, gender, 'race', schooling, ETS, frequency of foods containing earotenoids, total fat as a nutrient index, history of chronic lung diseases, and occupational exposure to asbestos. * Self-reported occupational exposure to any of: coal tar, formaldehyde, and chemicals. 0 o~ o~ o~ o o~

I I I 97 I I I I I I I I I I I I I I I 7.3 Main exposure variables 7.3 a. Self-reported exposure to ETS Table 31 shows the deaths from lung cancer among CPS II nonsmokers by any versus none self-reported ETS exposure. There was no indication of an association between self-reported ETS and the risk of lung cancer among nonsmokers when this exposure .variable was treated as dichotomous (i.e., any versus 0 hours of exposure to ETS ). The rate ratio adjusted for the age-gender distribution of person-time was 0.8 (95%CI--'0.6-1.0). The unadjusted (i.e., confounded by age) rate ratio was 0.3. Table 31 displays the lung cancer death rates in persons with no versus any exposure to ETS at home, work or elsewhere. The age-standardized rate ratio for men was 0.8 (95% CI=0.6-1.2) and for women 0.9 (95% CI---0.7-1.3). The age-adjusted Mantel-Haenszel mortality rate ratios were 0.6 (95% CI=0.4-1.0) and 0.9 (95% CI=0.7-1.2) for men and women respectively. The age-gender adjusted Mantel-Haenszel rate Patio was 0.8 (95%CI=0.6-1.0). 7.3.b Dose-response analyses of self-reported ETS Table 32 presents the analysis by tertiles of self-reported hours of exposure to ETS (i.e., 1-2 hours, 3-5 hours and 6+ hours). Panel A of-table 32 summarizes the information on the total number of persons in each category. This part of the table presents the number of lung cancer deaths, person years, and lung cancer death rates, among men, and women, separately and then combined in the last column. Panel B breaks down the previous numbers by five-year age groups. Thus, when the comparison of lung cancer mortality rates was made by duration of daily exposure to ETS between nonsmokers unexposed to ETS (i.e., no self-

98 I reported ETS exposure) and nonsmokers most heavily exposed (i.e., those in the upper tertile or exposed for 6 and more hours to ETS), subjects in this exposure category of ETS had a 20% increased risk of lung cancer, after adjustment for age and gender [RRm-h= 1.2 95% CI=(0.9-1.7)] Comparisons of the lung cancer death rates by terfile of ETS exposure among men and women are presented in Figure 8. Further adjustment of the association between ETS exposure and the risk of lung cancer was then conducted via Cox regression by blocking for age (12 five- year groups) 'race' (whites versus non-whites), schooling (<12 years of education, vs 12+ years), gender, asbestos exposure (ever versus never), and a history of chronic lung disease (any versus none) as dichotmous variables, 3 indicator variables for the intake of foods containing carotenoids (grouping those who had one or more a week into tertiles), and 4 other indicators for total fat as a nutrient index (grouping all subjects by quintiles). This coding of the c~variates is the same used to obtain the estimates presented in table 30. The results of the stratified Cox regression analyses are presented in table 33, and they show that inclusion of the covariates did not materially alter the reported association, once the confounding effect of age was controlled (Cfr. table 32 versus table 33). A multivariate test for dose-response was then conducted using this categorization of self-reported ETS as an ordinal variable, and failed to reject the null hypothesis. Separate analyses were conducted for the number of hours of exposure to ETS at different settings (home, work and other places), using the approach of simple and stratified analysis by age described for the cumulative hours of exposure to ETS. These results are summarized in table 33, along with the multivariate results for all ETS. The findings were the same as for the cumulative measure. I I I I I ! I I I I i I I I I I

Table 31. Lung cancer death rates among CPS II nonsmokers by self-reported ETS (any versus none), 1982-1989 Men Women Level: 0 hours of Any 0 hours of Any ETS ETS ETS ETS Age group Deaths P-Y Rate Deaths P-Y Rate Deaths P-Y Rate Deaths P-Y Rate <50 1 33,932 3 2 87,246 2 3 133,717 2 1 208,117 0 50-54 0 40,625 0 4 92,511 4 9 131,105 7 9 172,103 5 55-59 0 48,610 0 4 95,551 4 7 158,650 4 14 179,550 8 60-64 6 52,560 11 10 78,917 13 15 175,460 9 15 153,614 10 65-69 16 56,440 28 11 53,144 21 19 171,875 11 11 98,310 11 70-74 20 49,773 40 6 26,448 23 28 147,678 19 I I 48,843 23 75-79 11 31,781 35 3 11,025 27 40 105,983 38 4 23,445 17 80-84 15 15,519 97 2 3,803 53 32 59,479 54 3 10,389 29 85+ 5 7,368 68 0 1,273 0 22 35,831 61 3 5,937 51 ..... Tota! 74 336,614 14" 42 449,918 12" 175 1,119,778 10" 71 900,309 10" (n=329,226) *Age-standardized rates to the CPS II population. RR any versus none both men and women (age-gender adjusted)=0.8 (95% CI=0.6-1.0) RR any versus none for men (age adj. for men)=0.6 (95% CI=0.4-1.0) RR any versus none for women (age adj. for women)=0.9 (95% CI=0.7-1.2) LE90~9890~ 1 I N / l 99 l I

'Fable 32. A. (Summary) Lung cancer deaths, nonsmokers and person-years and rates by self-reported ..... ,, ETS category amoug nonsmokers in CPS I1 Men ..... Women Total Exposure Deaths Person- Rate* Deaths Person- Rate* Deaths Person- Rate* Years x10-5 Years x10-5 Years x10-5 All subjecls [ 16 110,688 786,532 16 246 281,538 2,020,081 I I 362 392,226 2,806,613 12 0 Iiours of E'PS 74 48,175 336,614 15 175 157,258 !,119,778 I ! 249 205,433 1,456,392 12 I-2 l louts of ETS 20 36,820 264,812 9 29 55,810 404,565 8 49 92,630 669,378 9 3-5 llours of ETS 8 10,301 74,060 18 I ! 24,255 175,376 9 19 34,556 249,436 I 1 ~...6+ HoursofETS 14 15,392 111,046 .!.7 31 44,215 320,362 12 45 59,607 431,408 13 Table 32". B. Age ~listribution and lu,n.g cancer deat~ rates by UnexpOsed ' Men ......... Unexposed "W6men (0 hours of ETS (0 hours of ETS ,_. , self-reported ETS category Unexposed Subjects (0 hours of ETS exposure) exposure) exposure) Age group Person- Rates Person- Rates Person- Rates in years Deaths years at (%) 10-5 Deaths years at (%) 10-5 Deaths years at (%) 10-5 dsk dsk risk <50 ! 33,932 10.1 3 3 133,717 11.9 2 4 167,648 ! 1.5 2 50-54 0 40,625 12.1 0 9 131,105 I 1.7 7 9 171,730 11.8 5 55-59 0 48,610 14.4 0 7 158,650 ! 4.2 4 7 207,260 14.2 3 60-64 -6 52,560 15.6 I 1 15 175,460 15.7 9 21 228,020 15.7 9 65-69 16 56,444 16.8 28 19 171,875 15.3 I ! 35 228,319 15.7 15 70-74 20 49,774 ! 4.8 40 28 147,678 ! 3.2 19 48 197,452 13.6 24 75-79 11 31,781 9.4 35 40 105,983 9.5 38 51 137,764 9.5 37 80-84 15 15,519 4.6 97 32 59,479 5.3 54 47 74,998 5.1 63 85+ 5 7,370 2.2 68 22 35,831 3.2 61 27 43,200 3.0 62 Total 74 336,614 100.0 ....... m'h age'adj Ri~ '-- ! '0 ' *Age-adjusted to the CPS-II population 15" 175 1,119,778 100.0 11" 249 1,456,391 100.0 12" m-h age-adj RR= i.0 m-h "gender-agd'-adj RR= 1.0 100

l ...... Ex~J0sed Men" (1-2 hours of ETS Table 32. (continued) Exposed Subjects (1-2 hours of ETS exposure) Person- Deaths years at risk 3 142,242 21.2 7 129,456 19.3 7 134,008 20.0 l0 ! 13,806 17.0 7 78,098 11.7 l0 41,397 6.2 4 19,341 2.9 0 7,687 1.1 1. 3,345 0.5 49 669,378 100.0 " ExpoSed Women (1-2 hours of ETS exposure) exposure) Age group Person- Rates Person- Rates in years Deaths years at (%) 10-5 Deaths years at (%) 10-5 risk risk <50 2 47,965 18. I 4 1 94,276 23.3 1 50-54 2 52,918 20.0 4 5 76,538 i 8.9 7 55-59 I 55,079 20.8 2 6 78,929 19.5 8 60-64 5 46,170 17.4 ! ! 5 67,636 16.7 7 65-69 3 33,230 ! 2.5 9 4 44,868 ! I. 1 9 70-74 5 18,087 6.8 28 5 23,309 5.8 21 75-79 2 7,800 2.9 26 2 11,542 2.9 17 80-84 0 2,713 1.0 0 0 4,974 1.2 0 85+ 0 851 0.3 0 1 2,494 0.6 40 Total 20 264,812 100.0 9* 29 404,565 I00.0 8* Rates (%) IO-5 2 5 5 9 9 24 21 0 30 9* Men: m-h age-adj. RR=0.6 (95%CI=0.4-1.0) Women:m-h age-adj. RR=0.8 (95%CI=0.5-1..2) Both: m-h gender-age adj. RR=0.7(95%CI=0.5- ! .0) *Age-adjusted to the CPS-II population 6890z9sgoE I I 101 l

mm m mmm mm mm m mm mm Table 32. (continued) Exposed Men (3-5 hours of ETS exposure) Age group Person- Rates in years Deaths years at (%) 10-5 Deaths risk <50 0 15,754 2 ! .3 0 0 50-54 0 15,635 2 !. 1 0 0 55-59 1 15,840 21.4 6 2 60-64 3 12,593 17.0 24 1 65-69 2 7,985 10.8 25 1 70-74 0 3,840 5.2 0 3 75-79 0 1,651 2.2 0 0 80-84 2 528 0.7 379 2 85+ 0 234 0.3 0 2 Total 8 74,060 100.0 18" 11 Exposed Women Exposed Subjects (3-5 hours of ETS (3-5 hours of ETS exposure) exposure) Person- Rates Person- Rates years at (%) 10-5 Deaths years at (%) 10-5 risk risk 39,517 22.5 0 0 55,271 22.2 0 33,808 19.3 0 0 49,443 19.8 0 34,675 19.8 6 3 50,515 20.3 6 28,857 16.5 3 4 41,450 16.6 I0 18,981 10.8 5 3 26,966 10.8 11 10,302 5.9 29 3 14,143 5.7 21 5,341 3.0 0 0 6,992 2.8 0 2,440 1.4 82 4 2,968 1.2 135 1,455 0.8 137 2 1,689 0.7 118 175,376 100.0 9* 19 249,436 100.0 II* m-h age-adj. RR=I.i (95%CI=0.5-2.3) m-h age-adj. RR=0.7 (95%CI=0.4-1.3) m-h gender-age adj. RR=0.9(95%CI=0.5-i.4) *Age-adjusted to the CPS-II population olzg0gg$90g 102 m~

103 Table 32. (continued) Exposed Men Exposed Women (6+ hours of ETS (6+ hours of ETS exposure) exposure) Age group Person- Rates Person- Rates in years Deaths years at (%) 10-5 Deaths years at (%)- I0-5 risk risk <50 0 23,527 2 ! .2 0 0 74,324 23.2 0 50-54 2 23,958 21.6 8 4 61,754 19.3 6 55-59 2 24,632 22.2 8 6 65,945 20.6 9 60-64 2 20,154 18.1 10 9 57,122 17.8 16 65-69 6 1 ! ,929 10.7 50 6 34,462 10.8 ! 7 70-74 ! 4,521 4.1 22 3 15,232 4.8 20 75-79 1 1,575 1.4 64 2 6,562 2.0 30 80-84 0 563 0.5 0 1 2,975 0.9 34 85+ 0 188 0.2 0 0 1,988 0.6 0 Total 14 111,046 100.0 17" 31 320,362 I00.0 12" Exposed Subjects (6+ hours-of ETS exposure) Person- Deaths years at (%) risk Rates 10-5 0 97,850 22.7 0 6 85,712 19.9 7 8 90,576 21.0 9 I 1 77,276 17.9 14 12 46,390 10.8 26 4 19,752 4.6 20 3 8,137 1.9 37 1 3,538 0.8 28 0 2,176 0.5 0 45 431,408 100.0 13" " m-h age-adj. RR=I.4 (95%CI=0.7-2.5) m-h age-adj. RR=I.2 (95%CI=0.8-!.8) m-h gender-age adj. RR= 1.2(95%CI---0.9-1.7) *Age-adjusted to the CPS-II population I~90~9~90E

I04 CP$ II, 1982-1989 a) 1-2 hours versus 0 hours 20 • Figure 8. Mortality rates from lung cancer among nonsmokers, by tertiles of self-reported ETS exposure and among unexposed, Rate per 100,000 population <50 50- 55- 60- 65- 70- 75- 80- 85+ 54 59 64 69 74 79 84 Age group ~=~ 1-2 hours hours I I I I I I I I I I I

I I I I I b) 3-5 hours versus 0 hours L40 ~ m Rate per 100t/ 80 100,000 50 population 40 2O 0 <~0 50- 55- ~ 65- 7~ 7~- 8~ 85+ 54 59 ~ 59 74 79 84 Age group ~~ 3-5 ho~ ET$ ¢ 0 ho~ c) 6+ hours vers~ 0 hours Rate per 50 I0 ,0 0 / ~ 10 ~ - ~0 5~ 55- ~ 65- 7~ 75- 8~ 85+ ~ 59 ~ 69 74 79 84 Age group ~"~ 6+ hours hours 105

106 None of the rate ratios by increasing amount of hours of self-reported ETS exposure at home, work or elsewhere, displayed in table 33, showed a statistically significant slope of a linear trend using Cox recession analysis. Moreover, there is a consistent pattern of risk deficit for low self-reported ETS exposure categories. Table 33. Cox regression multivariate lung cancer rate ratios for ETS exposure, cumulative and for specific sites, by ender, among CPS II nonsmoking participants, 1982-1989. Men Women Total Hours Multivariate Multivariate Multivariate of exposure Rate Rado§ Rate Ratio§ Rate Ratio§§ source_ (95 (95 (95 All ETS 0 1.0 1.0 1.0 1-2 0.6 (0.4-1.1) 0.8 (0.6-1.3) 0.7 (0.5-1.0) 3-5 1.0 (0.4-2.0) 0.7 (0~4-1.3) 0.8 (0.5-1.3) 6 + hours 1.3 (0.7-2.4) 1.1 (0.8-1.7) 1.2 (0.8-1.7) Home 0 1.0 1.0 1.0 1-3 0.7 (0.2-2.0) 0.4 (0.2-1.0) 0.5 (0.2-1.0) 4-5 0.0 (0.0-NC) 0.7 (0.3-1.7) 0.6 (0.2-1.4) 6 + hours 0.5 (0.1-3.9) 1.3 (0.8-2.1) 1.2 (0.7-1.9) Work 0 1.0 1.0 1.0 1 0.7 (0.3-t.6) 0.9 (0.5-t.9) 0.8 (0.5-I.4) 2-6 1.0 (0.5-2.1) 1.1 (0.6-2.1) 1.1 (0.6-1.7) 7+ hours 1.8 (0.9-3.6) 1.0 (0.5-1.8) 1.2 (0.8-2.0) Other places 0 I 2 3 + hours 1.0 1.0 1.0 0.5 (0.3-1.0) 1.0 (0.6-1.7) 0.7 (0.5-1.1) 0.7 (0.2-2.2) 0.8 (0.3-2.2) 0.7 (0.4-1.6) I. 1 (0.4-3.0) 1.1 (0.5-2.5) 1.I (0.6-2.0) § Adjusted for age, race, education, intake of carotenoid-containing foods, total fat as a nutrient index, occupational exposure to asbestos and history of chronic lung diseuse. §§ Additionally adjusted for gender. 0 0 ! I I I

107 I I I I I I I I I I I I I I I 7.3.c. Spousal ETS The relationship between environmental tobacco smoke exposure from spousal smoking and lung cancer mortality among nonsmokers was then assessed. Exposure to ETS based on exposure by a nonsmoking spouse to tobacco smoke from a smoking spouse was further defined based on whether the spouse was a nonsmoker or ever smoked, if the spouse was a current or former smoker (i.e., if the nonsmoker was ever, current or formerly exposed to ETS from the smoking habits of spouse). Comparisons of lung cancer death rates by ETS exposure from ever versus never smoking spouses showed no indication of an increased risk: the Mantel-Haenszel age-gender adjusted rate ratios were 1.0 (95% CI=0.7-1.4) for all nonsmoking spouses, and 0.9 (95% CI--0.7-1.5) for husbands and 1.1 (95% CI=0.8-1.5) for nonsmoking wives, respectively. Table 34 shows lung cancer mortality associated with exposure and no exposure to ETS from current smoking spouses for men and women, separately, and by the nine five-year age groups. The fu,'st four columns of table 34 present the data for nonsmokers married to current smokers of any type of tobacco product; the last four present the corresponding data for nonsmokers married to nonsmokers. Examination of lung cancer death rates presented as person-years for spouses, of either gender, shown in Table 34, are not appreciably different whether they were exposed or unexposed to ETS from a current smoking spouse. For example, the death rate columns for these ETS-exposed spouses show no appreciable differences across age groups among men, though a slightly greater mortality was observed among older women, as is graphically presented in figure 9. For nonsmoking spouses married to current smokers of any type of tobacco the RR was slightly above unity [ RR(m-h)=l.2 (95% CI--0.8-1.9)] for men and women combined, after adjusting for the age and gender distribution. For men, marriage to a current smoker was not associated with an increased risk [R_R(m-

108 I h)=0.9 95% CI-0.4-1.9)]. The corresponding age-adjusted estimate for nonsmoking women married to a current smoker was 1.3 (95% CI=0.8-1.9). Nonsmokers married to former smokers had no increased risk of lung cancer: the age-adjusted estimate for men and women combined was 1.0 (95% CI=0.7- 1.3). This was true both for men and women in our study, with age-adjusted rate ratios of 0.9 (95% CI---0.5-1.6), and 1.0 (95% CI=0.7-I.5), respectively. Most of smoking spouses smoked cigarettes. The relationships described above regarding current smoking spouses were true also for current cigarette smoking spouses: men married to current cigarette smoking women had an age-adjusted rate ratio of 0.9 (95%CI=0.3-1.9), whereas women married to current cigarette smokers had an age-adjusted rate ratio of 1.2 (95%CI=0.7-2.0). The age-gender adjusted RR was i. 1 (95% "CI=0.7-1.7). I I I I I ! I I I I I I I I

I I I I- I I I i I i I I I I I I I I I09 1aisle J4. Age-specific tung cancer rates among nonsmokers by smoking status (current §-any type of tobacco vs. life-long nonsmokers) of the spouses, CPS II, 1982-1989. a. Men " Deafl~s PY among De.~tbs ' PY ~nong ...... Age Group among married to (%) Rate among roan'led to (%) married to curr~nt 105 married to never cun~nt smokers never smokers smokers smokers <50 0 9,133 14 0 1 64,314 12 2 50-54 i 12,593 20 8 1 88,297 17 1 55-59 0 13,568 21 0 4 97,670 19 4 60-64 4 11,509 18 35 12 91,288 17 13 65-69 2 8,760 14 23 23 78,696 15 29 70-74 0 5,031 8 0 18 55,503 11 32 75-79 0 2,180 3 0 8 30,941 6 26 80-84 0 701 1 0 9 13,672 3 66 85+ 0 177 0 0 3 5,083 1 59 Total 7 63,652 100 10" 79 525,464 100 14" § Excludes current smoking spouses with/acomplete smoking data. Age adjusted MH RR~3.9 (95% CI=0.3-1.9). *Age-adjusted to the CPS II population b. Women Deaths PY among Deaths PY among Age Group among married to (~) Ram among married to (%) Rate marriedto curr~nt 105 married to never 105 current smokers never smokers smokers smokers <50 0 77,591 21 0 2 108,378 21 2 50-54 4 72,473 19 6 6 94,967 18 6 55-59 7 76,798 20 9 8 94,598 18 8 60-64 14 66,410 18 21 8 85,095 16 9 65-69 3 45,069 12 7 7 66,320 13 11 70-74 7 24,182 6 29 13 41,303 8 31 75-79 5 10,575 3 47 6 20,762 4 29 80-84 3 3,249 i 92 1 7,534 I 13 85+ 0 727 0 0 0 2,105 0 0 Total 43 377,074 100 15" 51 521,062 100 11" '~'Excia~des current smoking spouses with incomplete cigarette smoking data Age adjusted MH RR"-I.3 (95% CI=0.8-1.9). *Age-adjusted to the CPS II population

Figure 9. a) Rates of lung cancer among nonsmoking men by smoking status of their wives, CPS 11 1982-1989 35 Rate per 2~0 100,000 population <50 50-54 55-59 60-64 65-69 70-74 75-79 80+ Age group --'~ Wife Currently --X-- Wife never Smoked smoked Figure 9. b) Rates of lung cancer among nonsmoking women by smoking status of their husbands, CPS 11 1982-1989 80 Rate per population 0 -~m"~~ I I I - I I +I I <50 50-54 55-59 ~ 65-69 70-74 75-79 813+ Age group Husband Currently --X-- Husband Smoked Never S rooked I I I I I ! ! I I I

111 I I I I I I I I I I I I i I Table 35 presents lung cancer deaths, person-years and lung cancer death rates among nonsmoking women exposed to ETS from pipe/cigar current smoking of their spouses, compared to those among women married to nonsmokers. Death rates for lung cancer among nonsmoking women married to current pipe/cigar smokers increased more rapidly after age 70, than in those married to nonsmoking husbands. However, the small numbers of deaths make age specific estimates unstable. Nevertheless, it seems that exposure to ETS from spousal pipe/cigar has a weak statistically insignificant effect on the risk of lung cancer. More details on the types of smoking habits of spouses of nonsmoking subjects and their risk of lung cancer is displayed in table 36, along with a summary of the results of spousal ETS analysis described above. The ftrst row of table 36 presents the number of lung cancer deaths and person-years among nonsmokers in the entire cohort. The second row presents those numbers for men and women married to nonsmokers, as the referent category for analyses of the effect of ETS from spousal smoking. Thereafter those numbers for each category of smoking spouses are given along with age and age-gender adjusted Mantel- Haenszel rate ratios, as well as multivariate rate ratios controlling for all relevant covariatcs included in analyses of self-reported ETS. Nonsmoking wives married to current cigarette, pipe and cigar smokers showed an elevated risk of lung cancer. Nonsmokers married to former smokers, except for cigar/pipe smoking spouses, did not have an increased risk of lung cancer.

t12 Table 35. Age-specific lung cancer rates among nonsmoking women by cigar/pipe smoking status (current vs. life-time nonsmokers) of their husbands, CPS II, 1982-1989. Deaths PY Deaths PY Age among among (%) Rate among among (%) Rate Group married married 105 married married 105 tO tO tO tO current current never never pipe/cigar smokers smokers smokers smokers <50 0 24,967 17 0 2 108,378 50-54 0 25,615 18 0 6 94,967 55-59 0 28,490 20 0 8 94,598 60-64 7 26,281 18 27 8 85,095 65-69 2 I8,945 13 11 7 66,320 70-74 5 11,096 8 45 13 41,303 75-79 3 5,526 4 54 6 20,762 80-84 1 1,918 1 52 1 7,534 85+ 0 494 0 0 0 2,105 21 2 18 6 18 8 I6 9 13 I1 8 31 4 29 1 13 0 0 Total 18 143,341 I00 15" 51 521,062 100 11" ..... ~gg~djusted Mantel-Haenszel RR= 1.3 (95% CI'--2-0.7-2.2). *Age-adjusted to the CPS rr population This relationship between current smoking of spouses and the risk of lung cancer did not change when 'race', schooling, history of lung disease, frequency of consumption of foods containing carotenoids, and occupational asbestos exposure were allowed into the Cox regression model, along with age, as shown in table 36, suggesting that there was no confounding by these covariates. 0 I I I I I I I I I I i I

i I I- I I I I I I I I I I I I I I lahle 36. A~soctatton Oetween smoking status~' type ot smoking o3 spouses nonsmoking CPS II sub,iects, and lung cancer risk. Husbands Wives Total Spouse smoked tobacco Deaths Person-Years Deaths Person-Years Deaths Person-Years Total 10'1 '~ '7i9,044' ' i64 Never 79 525,464 5 [ RRm-h* 1.0 RRCox§ 1.0 Ever ** (any 22 193,580 113 type) 0.9 (0.5-1.5) RRm-h* 0.9 (0.6-1.4) RRCox§ Current (any 8 67,689 44 385,676 type) 0.9 (0.3-1.9) 1.3 (0.8-1.9) RRm-h* 1.0 (0.5-2.0) 1.3 (0.8-1.9) RRCox§ Former (any 13 117,462 68 614,96 i type) 0.9 (0.5-1.6) 1.0 (0.7-1.5) RRm-h* I. I (0.6-2.8) I. I (0.7-1.6) RgCox§ Ever 22 193,580 74 709,944 Cigarettes R~m-h* 0.9 (0.5-1.5) 1.0 (0.7-1.5) RgCox§ 1.0 (0.5-2.0) I. I (0.8-1.6) unclass, ever I 8,429 I 23,236 smoker Current 7 63,652 25 233,743 cigarettes incomplete I 4,037 I 8,592 RRm-h* 0.9 (0.4-1.9) 1.3 (0.8-1.9) RRCox§ 1.0 (0.5-2.0) 1.3 (0.8-2.0) For~er I0 I03,945 44 414,146 cigarettes incomplete 3 13,517 3 30,227 Rgm-h* 0.9 (0.5-1.6) 1.0 (0.6-1.5) RRCox§ I. I (0.6-2.8) 1.2 (0.8-1.8) E vet 39 313,929 Cigar/Pipes m-h* 1.2 (0.8-1.8) Cox§ I. I (0.8-1.6) Current 18 143,341 m-h* I..3 (0.7-2.2) Cox 1.5 (0.8-2.9) Former 21 170,588 m-h* I. I (0.7-1.9) cox§ 1.3 (0.6-2.8) (.=3 i~, I0~) *M-a: age ~just~ using Man~l-~iae~zel 1,544,935 265 2,263,979 521,062 130 1,046,526 1.0 1.0 1.0 1.0 1,023,873 135 1,217,453 L~ (0.8-L5) l.t (0.8-t.5) 1.1 (0.8-1.6) 1.0 (0.8-1.4) 52 453,365 1.2 (0.8-1.9) 1.2 (0.8-1.8) 81 732,423 t.O (0.7-1.3) 1.0 (0.% 1.4) 96 903,524 1. I (0.7-1.4) 1.0 (0.7-t.8) 2 31,665 32 297,395 2 12,629 1. l (0.7-1.5) 1.2 (0.8-1.9) 54 518,091 6 39 43,744 0.9 (0.7-1.3) 1. I (0.7-1.9) 313,929 § Cox: multivariate regression using the proportional hazard model to control for age, gender (for estimates listed in last column), 'race' and schooling, asbestos, history of chronic lung disease, consumption of foods containing carotenoids, and total fat as nutrient index. **Includes 2 ever smokers with unelassifiable smoking (i.e., former or current cigarette smokers).

7.3.d. Relationship between lung cancer death and ETS exposure by amount, duration, and both The dose-response relationship between lung cancer death and ETS exposure was assessed in a variety of ways. We f'trst examined amount of ETS exposure by measuring the number of packs of cigarettes smoked by current or former cigarette smoking spouses. We also examined the number of years nonsmokers were exposed to ETS from the smoking of their spouses. Finally, we examined both amount and duration using pack-years as a measure of cumulative ETS exposure. For these three analyses, we included data about subjects whose spouses were ever cigarette smokers, as the ETS-exposed group, and subjects whose spouses were never smokers, as the referent group. Table 37 shows lung cancer deaths per person-years for ETS-exposed spouses and multivariate rate ratios by packs of cigarettes smoked by their spouses compared with the referent group of non ETS-exposed spouses. These estimates are presenied by gender and then combined for subjects exposed to a spouse who was either a current or a former smoker. The upper panel of the table presents the data to compare the rates among nonsmokers married to current smoking spouses, and the lower for former smoking spouses; rate ratios are presented by categories of amount of smoking grouped by packs of cigarettes. The risk of lung cancer among men married to current cigarette smokers only increased among those who smoked less than one pack [P,.R=2.0 (95%CI--0.9- 4.4)] but the rate ratios decreased for the categories of heaviest cigarette smoking. Among women there was also an increased risk for those exposed to ETS from less than one pack of cigarette smoking, and declined among those married to current heavy smokers. No consistent linear trend with amount currently smoked by spouses was found. I I I I 0 O~

m m Table 37. Lung Cancer Adjusted Rate Ratios (95% CI) among nonsmoking spouses according, to the amount of cigarette smoked by spouses*, CPS II, 1982-1989. Exposure Husbands Wives All Nonsmokers Packs of cigarettes: Deaths/PY multivariate RR Deaths/PY multivariate RR multivariate RR by current smokers (95% CI) (95% CI) (95% CI) Nonsmokers 79/525,464 1.0 51/521,062 1.0 1.0 <I pack 7/28,923 2.0 8/61,820 !.4 i.6 (0.94.4) (0.6-2.9) (0.9-2.7) 1 - i.9 packs 0/29,756 0 15/126,087 i.4 1.0 (0-NC) (0.8-2.6) (0.6-1.8) 2+ packs 0/4.973 0 2/45,836 0.6 0.5 (0-NC) (0.1-2.3) (0.1-2.0) p test for trend 0.26 0.66 0.90 by former smokers Deaths/PY multivadate RR Deaths/PY multivariate RR multivariate RR (95% CI) (95% CI) (95% CI) Nonsmokers 79/525,464 1.0 5 !/521,062 ! .0 1.0 < ! pack 5/64,258 0.6 10/I08,365 0.8 0.8 (0.3-1.6) (0.5:1.8) (0.5-1.4) 1 - i.9 packs 4/32,191 1.0 20/213,304 0.8 0.9 (0.3-2.8) (0.5-1.4) (0.6- !.4) 2+ packs 1/7,495 1.2 14/92,462 1.5 1.5 (0.2-1.9) (0.8-2.7) (0.8-2.6) p test for trend 0.74 0.58 0.72 a '--258,499) ~Only cigarette smokers (current and former) with complete data. Cox regressio.n model stratified for age, gender, 'race', schooling, total intake of foods containing carotenoids, total fat ntake, occupational exposure to asbestos and history of chronic lung disease. 8RgOE9890E m n m m m m m m m mm m m 115 mm ~m

I16 I I I I ! A possible problem with the aforementioned analysis is that some ETS-exposed spouses may have been previously married to someone who was not a smoker. Thus, it was decided to re-examine the realtionship between lung cancer mortality and ETS exposure by packs of cigarrettes smoked by spouses by restricting to an analysis of spouses married only once in their lifetime. Table 38 presents limg cancer deaths, person years and multvariate rate ratios by amount of cigarettes smoked by current or former smokers as was presented above for the full data set. The analysis is restricted to the 148,402 spouses married once and who had complete information on age at marriage. The same group of nonsmokers unexposed to ETS is the referent. Unlike in the previous analysis of the fulI data set of nonsmokers married to cigarette smokers, in this subset of spouses married once in their life-time, among nonsmokers married to former smokers, there is a slightly increased risk of lung eancer for those married to former smokers who smoked 2+ packs of cigarettes. However, there is no statistically significant trend: the p value of multivariate Cox regression analyses of the packs of cigarettes smoked by former smokers were 0.28 for men -decreasing trend, and 0.29 for women whose risks showed an increasing, but inconsistent trend, and 0.6 for both men and women. 0 O~ O~ 0 O~ I I I / I I

[___ l Table 38. Lung Cancer Adjusted Rate Ratios (95% CI) among nonsmoking spouses according to the amount of cigarette smoked by spouses* if married once and with data on age at marriage, CPS II, 1982-1989. Exposure Husbands Wives All Nonsmokers Packs of cigarettes: Deaths/PY multivariate RR Deaths/PY multivariate RR multivariate RR b~, current smokers (95% CI) (95% CI) (95% CI) Nonsmokers 46/314,944 1.0 30/31 ! ,333 i .0 i .0 < 1 pack 5/14,310 3.0 5/32,524 1.7 2.1 (1. 1-7.9) (0.7-4.4) ( !. 1-4.1 ) i - !.9 packs 0/15,054 0.0 10/69,060 1.6 !.2 (0.0-NC) (0.8-3.4) (0.6-2.3) 2+ packs 0/2,308 0.0 2/24,900 0.9 0.8 .... (0.0-NC) (0.2-3.9) (0..2-3.5) p test for trend 0.6 0.34 0.55 by former smokers Deaths/PY multivariate RR Deaths/PY multivariate RR multivariate RR (95% CI) (95% CI) (95% CI) Nonsmokers 46/314,944 ! .0 30/311,333 i .0 1.0 < I pack 1/34,042 0.2 4/61,677 0.6 0.5 (0.0-117) (0.2-1.8) (0.2-1.2) ! - !.9 packs 0/15,915 0,0 12/120,585 0.8 0.7 (O.O-NC) (0.4- 1.7) (0.3-1.4) 2+ packs i/3,559 2.8 i !/49,304 2.0 1.9 (0.4-21.6) (1.0-4.0) (1.0-3.7) p test for trend 0.28 0.29 0.6 (n=148,204) "" * Analyses restricted to nonsmoking spouses married to nonsmoking spouses and those married to cigarette smokers (and not other type of tobacco), with complete smoking data, married once at the time of interview, and with valid data on age at first marriage. § Cox regression model stratified for age, gender, 'race', schooling, total intake of foods containing earotenoids, total fat as nutrient index, occupational exposure to asbestos and histozy of chronic.lung disease. NC=not calculable l 1 I i I l l I l I I !!7 ml

118 The hypothesis of this study was submitted to a more severe test to take into account the time these spouses married once in their life-time spent together (i.e., whether or not these nonsmokers were exposed to ETS from the smoking habits of their spouses). The relationship between lung c;incer mortality and ETS exposure by duration was examined. To conduct these analyses, it was necessary to estimate the number of years nonsmokers were exposed to ETS from spousal cigarette smoking. For this variable, the referent groups represented spouses who were not exposed, either because thay were married to nonsmokers or to former smokers who quit smoking before marriage (i.e., in doing so, those nonsmokers were never exposed to the tobacco smoke of their spouses). This resulted in reclassifying from the exposed categories 4 percent of the person-time, and 1 death (0.8 percent) in this analytic cohort, to the unexposed category. Therefore, the specificity of classification of exposure to ETS was increased. Table 39 shows deaths per person-years for ETS- exposed spouses by duration, accounting for the eight covariates presented in table 30. Distribution of time in marriage are gender-specific. For estimates of the RR for both men and women, we used the combined distribution of time in marriage to smokers. Nonsmoking men married to smokers for 15 or more years did not have an increased risk of lung cancer, although there were fewer persons in these categories. Nonsmokers married up to 15 years to smoking wives had a 30 percent increased risk. We found no evidence that the rate ratios increased among nonsmoking men by time in marriage with smokers. However, rate ratios increased among women as the time in marriage from one to seventeen years to smokers, and then decreased slightly, in an erratic trend shown in table 39. | I I I I I ! I I I I I !

i i i Table 39. Lung Cancer Adjusted Rate Ratios (95% CI) among nonsmoking spouses according to time in marriage with current cigarette smoking spouses*, CPS II, 1982-1989. Hd~bafids Wives Both" Years married Deaths/PY mul[ivariate Years ""' ' Deaths/PY multivariate Years ..... "multivariate to smoker RR married to RR married to RR (95% CI) smoker (95% CI) smoker (95% CI) 0 years 46/329,905 1.0 0 years 30/334,946 1.0 0 years 1.0 (Nonsmokers and (Nonsmokers (Nonsmokers quittes before and quitters and quitters marriage) before marriage) belbre marriage) 1-15 1/22,101 0.4 1-17 13/107,681 1.5 1-17 1.2 (0.0-3.1) (0.8-3.0) (0.6-2.2) 1/29,918 1.2 14/112,761 1.3 18-29 i .2 (0.0-3.7) (0.7-2.6) (0.7-2.2) 5/18,208 0.7 ! 7/114,002 1.2 30+ 1.0 (0.2-2.7) (0.6-2.2) (0.6- !.8) 16-26 18-29 27+ 30+ test for trend 0.76 test for trend 0.49 test for trend 0.72 (n=148,204) * Analyses restricted to nonsmoking spouses married to nonsmoking spouses and those married to cigarette smokers (and not other type of tobacco), with complete smoking data, married once at the time of interview, and with valid data on age at first marriage. § Cox regression model stratified for age, gender, 'race', schooling, total intake of foods containing carotenoids, total fat as nutrient index, occupational exposure t.o asbestos and history of chronic lung disease. LS9OE9EgoE

120 Finally, the relationship between lung cancer deaths and ETS exposure was examined by both amount and duration. Pack-years, the exposure variable for this analysis, was created by multiplying the packs of cigarettes (i.e., amount) by the number of years of exposure (i.e., duration). For example 20 pack-years could have been reached during marriage for 20 years with a smoker who smoked one pack of cigarettes daily. This variable represents cumulative exposure to ETS over time. As mentioned above for time in marriage, for estimates of the RR for both men and women, we used the combined distribution of pack-years of smoking spousal. Table 40 presents lung cancer deaths, person-years, and rate ratios among nonsmoking men, women and then both men and women, by pack-years according to the quintiles of the distribution of pack-years of smoking of the spouses during marriage with further adjustment for the same confounders in table 30. Nonsmoking husbands were exposed to considerably less ETS from spousal smoking than nonsmoking wives. As was shown before fewer nonsmoking men fell into any of the categories of heavy spousal ETS from cigarette smoking (i.e., 15+ pack years), whereas, 32 percent of the nonsmoking women experienced such exposure and were evenly divided across the categories of pack-years of cigarette smoking of their cigarette smoking husbands. The multivariate rate ratios of lung cancer among nonsmoking men increased by cumulative exposure to ETS up to 22 pack-years of cigarette smoking, and then decreased. Thus no consistency in the variation of lung cancer risk and this measure of long-term ETS exposure was found among nonsmoking men. However, among women, the rate ratios increased consistenly by pack-years of cigarette smoking of theh" husbands, from 1.1 among slightly ETS exposed women, to 1. i among women exposed from 17 to 35 pack-years, and then roughly reached a 50% increased risk for women exposed from 36 pack-years and more. The associated p value for the multivariate test of linear trend was 0.14, thus failing to reject the null hypothesis of nolinear trend. r~ o 0~ I I I I I | I i I I I l I

121 | I- I I I I I I i I I I I I I I Pack-years of cigarette smoking during marriage, was not statistically significant associated with increasing risk of lung cancer of both nonsmoking husbands and wives (p=0.54) (Table 4-0).

Pack-years Table 40. Lung Cancer Adjusted Rate Ratios (95% CI) among nonsmoking spouses according to pack-Tears of spousal cigarette smoking, CPS H, 1982-1989. Husbands Wives Both Dcaths/PY multivaJ-iate'RR " Deaths/PY multivariate RR multivariate RR 0 46/329,905 ! -8 1/24,018 9-22 2/23.438 23+ 2/23,862 (95% CI) Pack-years (95% CI) (95% CI) 1.0 0 301334,946 ! .0 1.0 (0.1-2.9) (0.5-2.2) (0.6- 1.9) !.4 17-35 16/113,119 1.3 !.2 (0.5-4.2) (0.7-2.5) ~0.7-2. I) 0.5 36+ 18/IO9,0o6 1.5 I. ! (0, I-2,2) (0.8-2.8) (0.6- !.9) test for trend 0.54 test for trend p=0.14 p=0.54 (n= 148,204) * Analyses restricted to nonsmoking spouses married to nonsmoking spouses and those married to cigarette smokers (and not other type of tobacco), with complete smoking data, married once at the time of interview, and with valid data on age at first marriage. § Cox regression model stratified for age, gender, 'race', schooling, total intake of foods containing earotenoids, total fat as nutrient index, occupational exposure to asbestos and history of chronic lung disease. 122 0990~9~90~ Ill -| Ic- | II~ II I I-"| | | |

i I I 123 I I I I I I I I I I I I I I 7.4 Joint effects of ETS and asbestos exposure Further analyses were conducted to describe the effect of ETS among those subjects exposed to asbestos. Occupational exposure to asbestos was reported by only 2.5 percent of CPS II participants. As shown in Table 2I, asbestos exposure was reported three times more frequently by men than women. Using the CPS II cohort for analyses of self-reported ETS, we contrasted lung cancer death rates of grouping subjects by tertiles of self-reported ETS exposure and by ever or never exposed to asbestos. Nonsmokers heavily exposed to ETS ('e_6 hours) in 1982, and who had ever been exposed to asbestos at work, experienced a higher risk of lung cancer than expected if the effects of ETS and asbestos were independent. Table 41 shows the results of these analysis. The formal test for interaction in the multiplicative scale using the Cox regression model with both asbestos and ETS exposure variables (reduced model), controlling for age, schooling, gender, 'race', consumption of foods containing earotenoids, total fat as nutrient index, and history of lung disease, yielded a -2 In likelihood of 4125.88, and that with asbestos and ETS and the three interaction terms of asbestos and tertiles of ETS exposure (full model) was 4121.002, for a LRz3~,2=4.878 with an associated p value of 0.18. Nonsmoking men and women who reported 6 or more hours of exposure to ETS, and ever being exposed to asbestos had four times the risk of lung cancer [multivariate RR=4.5 (95%CI=0.4-48.7)] compared to those of nonsmoking CPS II participants who had neither of those environmental exposures. The multivariate association with asbestos alone was RR=I.5 (95%CI----0.7-3.2). I~) 0 0-~ O' O~

I I Table 41. Rate ratios (95% CI) for lung cancer among CPS II nons]noklng men and women according to exposure to ETS (cumulative at home, work, and elsewhere) and occupational exposure to asbestos. Exposure to ETS +6 hours 3-5 hours !-2 hours 0 hours Deaths Person-Years Deaths Person-Years Deaths Person-Years Person-Years at at risk at risk at risk Deaths risk Asbestos Yes 5 14,954 0 7,922 2 22,250 7 Rate* 33 x 10-5 0 11 x 10-5 24,436 16 x 10-5 RR 4.5 0 0.9 1.5 Cox§§ (0.4-48.7) (0.1-11.1) (0.7-3.2) Asbestos No 40 416,453 19 241,513 47 647,134 242 !,431,956 Rate* 13x10-5 llxl0-5 9x10-5 !1 x 10-5 RR Cox§ !. 1 0.8 0.8 1.0 (0.8-1.6) (0.5-1.4) (0.5-1.0) =Age adjusted to tile distribution of the CPS 11 population ~ Cox regression model stratified for age, gender, 'race', schooling, history of chronic lung disease, frequency of consumption of foods containing carotenoids, and total fat intake. ~990Z9~90~ .I 124 I

125 I I I I I I I I I I I I I I 7.5 Model Specification Proportional hazards modeling was the main analytic tool used in this study. Therefore, a valid question to ask is whether the proportional hazard assumption held. Univariate survival curves using Kaplan-Meier estimates for age, gender, • schooling, 'race', consumption of foods containing carotenois, and total fat, history of chronic lung disease, and occupational exposure to asbestos, all followed a pattern of parallel curves by follow-up time in CPS II. Since most analyses on ETS (either self-reported exposure or from spousal smoking) were. conducted while blocking for the covariates, we present univadate Kaplan-Meier estimates of survival for the main exposure variables themselves, displayed in figure 10.

126 I Figure 10. Log I-log(S)] curves for grouped data analyses of A. self-reported ETS Ll~(,.L~f(ka'vlvtl }t~lll~)| |i~lll|is -I| * A=unexposed, B=1-2 hours, C=3-5 hours, and D= 6+ hours of exposure to ETS. I I I I I I I I i I I

I I I 127 I I I I I I I I I I I I I I Figure 10 B. ETS from pack-years of spousal smoking L 'i | -I0 ' =1.I , 1 t~U.~UP Tile k~J LICA A---O, B=<7 pack-years, C=7-14 pack-years, and D= 15+ pack-years.

128 I We also conducted Poisson regTession analyses, an alternative choice of the Cox regression model. RR estimates from the Poisson model for self-reported ETS (upper tertile > 6 hours of ETS) and current spousal ETS from cigarette smokers were 1.2 (95% CI= 0.8-1.6) and 1.2 (95% CI----0.9-1.8), respectively. Thus, the general results from Poisson regression modeling, closely agree with those presented using the Cox regression model. 7.5 Leading causes of death in the cohort During the same period of follow-up among the 314, 108 nonsmoking participants in the the spousal ETS analytic cohort, there were 12,792 other deaths. Coronary heart disease was the leading cause of death in this g'roup, with 3,742 deaths (29.2 percent). The major causes of death according to ICD-9 codes are displayed in table 42. Table 42. Number of deaths from major smoking-related c__auses__a_m_0_ng nonsm0king spouses in the CPS II, 198.2-1989. Causes of death (ICD-9) Deaths % Ischemic heart disease (410-414) Stroke (430-438) Upper aerodigestive cancer -mouth, pharynx, larynx, and esophagus (140-150, 161) Other cancers (140-209) Lung cancer (162) Duodenal or gastric ulcer (531-534) Cirrhosis and alcoholism (57 I, 291,303) Hypertensive heart disease (401-405) Injuries ('E810-E988) Other medical causes (000-799) All causes 3,742 29.2 724 5.6 36 0.3 909 7.1 265 2.1 25 0.2 116 0.9 120 0.9 569 4.4 6,286 49. i 12,792 100.0 I I I I I I I I I I I i I

129 I I" I I I I I I I I I I I I I I Chapter 8: Discussion and Conclusions 8.1 Consistency Since t98 t, when the first study that examined the relationsbAp between ETS and lung cancer death was published (I-Iirayama, 1981), 35 other studies that have examined this same relationship have beech published (Appendix A and Table 2). Of these 36 studies, four are cohort studies and 32 ease-control. It is known that in case-control studies, in which the "information most times comes from the subject of proxy respondents after disease onset, knowledge of the disease could affect exposure data" (i.e., introducing a recall bias) (Rothman 1986). Cohort studies are less subject to recall bias and therefore lend themselves more than case-control studies to making inferences about cause and effect. Thus, although many published studies are available, only a few can be considered to have assessed the relationship between ETS and lung cancer risk in such a way that the measurement of ETS exposure preceds the occurrence of lung cancer. Of those four cohort studies on ETS and lung cancer, one included eight lung cancer cases (Butler 1988), another had nine (I-Iole t989), a third had 153 (Garfinkel 1981), and the largest had 174 lung cancer deaths (I-Iirayama 1981). Our study is the largest cohort study to assess the relationship between ETS and lung cancer death. Therefore, in this paper we report findings from the largest cohort study that are consistent with aggregated evidence that supports the existence of a relationship between cumulative ETS exposure and the risk of lung cancer among nonsmokers. This study makes use of a measure of exposure that combines duration and amount of exposure to ETS that had not been used before in previous cohort studies about the effect of ETS on lung cancer risk. This cumulative exposure to ETS, which is referred to as ETS exposure from pack-years of cigarette smoking of the spouse (Fontham 1991), attempts to estimate ETS long-term exposure. Because 90% of smokers smoke at home (1988 NHIS-OH, Table 1), spouses married to smokers are likely to be exposed to ETS in the home. Our measure of exposure reflects intensity and duration of exoosure to ETS during marriage, and may provide a more adequate measure of long-term ETS exposure. Therefore, this measure of exposure enabled us I

130 to estimate lung cancer risks a~bociated with increasingly greater ET3 exposure with regard to duration and amount. For example, in this study we found increasing lung cancer risks with increasing ETS exposure, with a 50% increased risk, although not statistically signficant, for the most exposed group versus those who were not exposed. In this study, we also found that this not statistically signficiant increased lung cancer risks associated with ETS exposure remained even after we adjusted for the effects of potentially confounding variables by means of Cox proportional hazards modeling. Most previously published studies that had examined the relationship between ETS exposure and lung cancer risk had not accounted for the effects of most known potentially confounding variables included in our models. Thus, questions had been raised about the possibility of spurious findings in past studies (Mantel 1992). In our study, we controlled for the effects of age, gender, socioeconomic status, race/ethnicity, fruit and vegetable intake, fat intake, occupational exposure to asbestos, and a history of chronic lung disease, and we still found that ETS exposure from pack- years of spousal smoking increased the risk of lung cancer. Therefore, our findings support the notion that the observed relationship is not the result of known confounding variables. All cohort studies on this issue have been based on lung cancer diagnosis from death certificates. As previously reported for lung cancer in the US, this approach provides a valid diagnostic tool for epidemiologic research (Percy 198 I). None of the previous studies verified their death certificate diagnoses with histopathologic data. Some have reviewed hospital records, and in one large case-control study histopathological slides were reviewed (Fontham 199 l). In our study, we verified death certificate diagnoses with cancer registry diagnoses on a 10 percent sample of lung cancer deaths (i.e., those of residents of SEER cancer registries areas). Most SEER cancer registry diagnoses (92%) are histopathologically confirmed (NCI-SEER t989). The proportion of the study subjects who died from lung cancer and resided in SEER cancer registries' areas, who were histologically confirmed was 86.2 percent. Seventy percent of all lung cancer deaths were adenocarcinomas. Thus, cases in our study are likely to have been primary lung cancer, and most were adenocarcinomas. ! I I l l I I I ,I I ! I '1 I ~0 0 (

131 I I" I I I l I I I I I I I I I I Adenocarcinomas are the specific histological type of lung cancer seen most often among nonsmokers. Although the information on specific histologic types was limited in our study, based upon the estimate of seventy percent of adenocarcinomas among the lung cancer deaths of nonsmokers, our findings of this study lend support to the hypothesis that a richer composition of SS in volatile carcinogen components more likely to reach the periphery of the lung would actually be responsible for the higher proportion of adenocareinomas among nonsmokers (Wynder 1983, Fontham 1991). Our findings on the association between ETS exposure from spousal smoking-and the risk of lung cancer agree with the combined estimate from 36 published studies, reporting a 20% increased risk of lung cancer among nonsmokers associated with this measure of exposure to ETS. In 1981 Garfinkel published the results of the second large prospective cohort study sponsored by the American Cancer Society, the Cancer Prevention Study I. This study.comprised a cohort of 1,078,894 men and women followed from 1960 to 1972. The CPS I analyses based on 153 Iung cancer deaths among nonsmoking women found a 20% percent increased risk of lung cancer, although this elevated risk was not statistically significant (95% CI=0.9-1.4) (Garfinkel 1981). Analysis of CPS I and CPS II agree in the magnitude of the effect of spousal smoking. As in most previous epidemiologic studies of ETS, we found a trend in the risk of lung cancer among nonsmoking wives with increasing levels of smoking by the husbands, although it was not statistically signficant. In contrast, for self-reported ETS we found no statistically significant evidence of an elevated risk among the ETS exposed individuals at interview. A case-control study by Kabat and Wynder found an association for self-reported ETS at work among men [3.1 (95% CI=I. 1-11.0)], but not women. In another case-control study that used self-reported ETS as one measure of exposure, Garfinkel et al. found no increasing trend with increasing exposure to ETS measured as number of hours of exposed to the smoke of others in the past, and the risk of lung cancer among nonsmoking women. This fact led the authors to conclude that "the lack of relationship when exposure was classified by

[32 .... hour~ exposed to smoke oL" others may have occurred because this variable does not accuratelv measure intensity of exposure". In this study, however, a two-fold increased risk was found for women whose husbands in the past smoked 20 or more cigarettes at home (Garfinkel 1985). Brownson et al. reported a 1.7 odds ratio for lung cancer for nonsmokers who had four or more hours of self-reported exposure to ETS in a small study of adenocarcinomas (Brownson 1987) but were unable to replicate their findings in a larger study (Brownson 1992). The negative findings of this study with respect to self-reported ETS exposure may well be due to misclassification of exposure since the questionnaire data on self- reported number of hours of exposure to ETS may reflect only current exposure and not the more biologically relevant past exposure. An important evidence of the possibility of such misclassification can be found in the decreased risk of persons in low self-reported ETS exposure categories. This might be due to the inclusion in the referent category (i.e., "0" hours or unexposed) of false negative unexposed persons particularly among those ,dith missing data on self-reported ETS exposure. Exposure to tobacco smoke from the spouse, as was measured in our study (i.e., self- reported smoking history of the spouse) probably provides a more reliable index of long-term and meaningful ETS exposure than current self-report ETS. This measure is not affected by dramatic changes in the prevalence of smoking seen since the 1960's in the US. It ensures that the smoker has a close relationship with the nonsmoker (i.e., spouse). Moreover, our measure of time in marriage takes into account the effective time spent with the smoker during marriage in such a way that if a smoker had quit smoking before marriage, nonsmokers were classified as unexposed to spousal smoking. By the same token, the time smoking spouses smoked in marriage was estimated to take into account the time since quitters stopped smoking. Our findings are generally consistent with those of other epidemiologic studies. Some case-control studies found an association with number of cigarettes or other measures of quantity usually smoked by husbands, but not with duration of spousal smoking (i.e., time living with a smoking spouse) (Hirayama 1984, Akiba 1986, Dalager 1986, Lam 1987, Inoue 1988), while the reverse was observed in some other epidemiologic ~0 O: O~ 0 O~ 0 I ! l ! I R

! I I I- I I I I I I I I I I I I I I I 133 studies tGao i987, Kaiandidi i990). The use of a combination of quantity and duration has been found associated in at least one other epidemiologic study (Fontham 1991). That study conducted by Fontham et al., is by far the best in this regard. The observations regarding the eft'ect of ETS among nonsmokers exposed to asbestos are consistent with those of other large epidemiologic studies which concluded that active smoking and asbestos act synergistically (Selikoff t968). This finding provides additional evidence in support of a causal relationship between ETS and lung carlcer. Some limitations of this study, such as statistical power and misclassification bias, are reviewed. 8.2 Study power The most obvious limitation of this study, shared with most other epidemiologic studies which have addressed this hypothesis, is limited power to detect with sufficient precision a RR on the order of 1.2 (i.e., the summary effect of ETS from 36 other studies). The power of the CPS II was approximately .50 percent for detecting this magnitude of association, as shown in table 43. If indeed ETS increases the risk of lung cancer among nonsmokers by less than 20 percent (e.g., I0 percent), then the power of this study to detect such association with sufficient precision would be only 20 percent. 0

[34 Table 43. Results of power calculations (binomial parameter p=0.5, the proportion of CPS II nonsmoking participants exposed to ETS, by M1 the total number of lung cancer deaths among nonsmokers Mortality Rate Po~') RatioSelf-reported ETS M 1=362 ........... 1.4 94. I 1.3 80.5 1.2 54. I 1. I 23.5 Spousal ETS M1=265 87.1 70.3 45.4 20.5 Thus, the lack of statistical significance in most analyses may reflect low statistical power. 8.3 Misclassification of self-reported ETS Misclassification of both self-reported and spousal ETS exposure might have affected the results of our study yielding a bias towards the null. A dilution of the effect from non-differential misclassification would obscure a weak association between ETS and lung cancer. If any misclassification occurred, probably it was non-differential (i.e., subjects who died from lung cancer were as likely to have misclassified themselves with respect to ETS exposure, as those who did not). Table 44 displays the results of using values in the range of 0.75-0.95 for specificity and sensitivity of classification of outcome or exposure variables in standard formulas (Kleinbaum 1982) to correct for misclassification of an observed association of 1.2 as observed in this study (Cf. Table 34, for nonsmoking wives comparing those married to nonsmokers versus those married to current smoking spouses). Each one of the parameters assumes the values in the x axis, while the others are assumed to have perfect validity. A meaningful adjustment for misclassification of ETS exposure would be necessary in the likely case of having classified exposure with a specificity below 90 percent. ! I ! I ! I I I I I I

i I I I- I I I I i I i I I I I i I I I 135 Table 44. Corrected RR from an observed value of 1.2 (using data from table 34, nonsmoking wives exposed to ETS from current spousal smoking), by degree of misclassification of ETS exposure{} Value of Specificity of Sensitivity of Exposure Parameer , Exposure .... 0.95 1.2 1.2 0.9 1.2 1.2 0.85 1.2 1.2 0.8 1.3 1.2 0.75 1.3 1.2 {}"Each p~ter ~h~ffes, while the c~er is held constan~ at 1 (i.e., perfect classification). In general subjects may have misclassified themselves with respect to their exposure status for the reasons discussed in 8.1. In addition, both refer to exposure during adulthood and thus do not take into account exposure during childhood. However, Fontham et al. study findings (1991) suggest that this source of bias probably is unimportant. Misclassification of relevant exposure to ETS, however, is more likely to have occurred for self-reported ETS than for .spousal smoking for the following five reasons. First, a large proportion of blanks in the CPS ]I ETS questionnaire section were interpreted as unexposed; this assumption may be unrealistic and therefore, augmented a dilution bias. Results of analyses restricted to those who had ftlled the three spaces provided for self-reported hours of ETS exposure (Table 45) showed that such dilution bias existed: the point estimate of the rate ratio of subjects with 6 and more hours of ETS exposure was 1.8 (95% CI---0.9-3.6). The rate ratio was found to be diluted upon inclusion of people with any blank for ETS, because when only those who left the three spaces blank were excluded, the rate ratio was 1.2 for those who were exposed for 6 or more hours to ETS. However, in that case the study had been conducted based on fewer deaths (i.e., 104, or 243, respectively), and therefore, would have had even less power. Those who left any space blank in the spaces provided to write down 0 O~ 0 O~

136 .... the number of hours exposed to E¥~ (and grouped with those who annotated zero hours) or had unquantifiable ETS data (and excluded from analysis) had a multivariate RR of 1.0 (95% CI=0.8-1.3) and 1.0 (95% CI=0.8-1.4), respectively, when compared with the rates of those who annotated zero hours in the three spaces. Table 45. Missings are Exposure to Unexposed ETS (n=362 deaths) Rate ratios § from ETS by different approaches dealing_with missing inform~atio~.on ETS Excludes Excludes ' missings in missings in all three any of fields three (n=243 fields deaths) (n=I04 deaths) 0 1.0 1.0 1.0 1-2 0.7 0.7 0.7 (0.5-1.0) (0.5- i. I) (0.4-1.2) 3-5 0.8 0.7 0.3 (0.5-1.2) (0.5-1.3) (0.1-1.9) 6+ hours 1.2 1.2 1.8 (0.8-1.7) (0.8-1.8) (0.9-3.6) § Adjusted for age, gender, race, education, intake of carotenoid-containing foods, total fat as a nutrient index, occupational exposure to asbestos and history of chronic lung disease. Second, a positive association between schooling with self-reported ETS, could be interpreted as proof of 'increased sensitivity' to the smoke of others among nonsmokers of higher SES. Based upon data from the 1988 NHIS-OH, and most other Smaller surveys and studies of ETS, we expec~ted to find that CPS II participants of relatively lower SES would have reported more ETS than those in. higher SES. Tables 21 and 22 show that there was a direct relationship between any self-reported exposure to ETS and years of education (i.e., higher educated participants reporting more exposure), whereas the opposite occurred with any spousal ETS. In table 46 we compared formal education with self-reported ETS status, and spousal smoking status (any versus none). For simplicity we restricted the comparison to the extremes of less than high school and college 0 0 ! I ! I ! I i I .I I I I ! I

I I I I I I I I I I I I I I I I I I i37 ~graduates and graduate school The resuits of this comparison clearly indicate that nonsmokers with higher education were more likely to report any ETS exposure, but less likely to be married to ever smokers, and suggests that self- reported ETS does not accurately reflect ETS exposure. Table 46. Comparison of any ETS exposure (self-reported or from spousal smoking) by years of education among CPS II nonsmokers, 1982. 16+ 62,731 56,10I 1.3 46,036 52,62I 0.9 Third, self-reported number of hours of ETS exposure does not necessarily reflect the intensity of ETS exposure, but duration to an undetermined amount of ETS. This limitation might contribute considerable misclassification of self- reported ETS. Fourth, as mentioned in 8.1, an indication that suggests such misclassifieation of ETS exposure is found in the results of self-reported ETS exposure itself. Unlike most spousal smoking analyses, there is a consistent pattern of deficit in the risk of lung cancer for the first and second tertile of self-reported ETS exposed, whether it is cumulative in the three settings, ETS at home, work or other places separately. These results are compatible with misclassification of an undetermined proportion of exposed who left blank spaces for hours of ETS exposure blank in the CPS II questionnaire. Last, the classification by self-reported ETS in 1982 has another inherent source of misclassification: that from the changing patterns of smoking (e.g., unexposed subjects in 1982 might have been exposed before if married to former smokers). Therefore, self-reported current ETS exposure does not assess long-term exposure, whereas smoking status of spouses might reflect exposure for many

i38 years, although it may still have iimited vaiidity (Garfinkei i98i, Frieaman I983). An estimate of the RR of lung cancer from ETS corrected for this downward misclassification bias, using the data available in NRC 1987 report which was RR=I.9 (Garm 1988). The EPA report included a correction for this bias (EPA 1992), and the pooled estimates were in the order of our metanalysis estimate (i.e., 1.2). Correcting for this downward bias would result in estimates of 1.2. GarfinkeI et al. pointed out, when reporting the findings of CPS I, that "Long- term effects of passive smoking are difficult to establish because of the problems of classification. It may be misleading to classify a woman as a passive smoker or not on the basis of her husband's smoking habit. Wives of nonsmokers may be more exposed to cigarette smoke of others than wives of cigarette smoking men; wives of smokers may be very little exposed to the cigarette smoke of their husbands or other" (Garfmkel 198 I). In the hypothetical situation of randomly misclassifying 10-25 percent of the study participants, any bias is towards the null: thetrue effect of ETS would be at least as great as the point estimate, and the size of the bias would range from -0.03 to -0.4. Notice that in the typical stituation the bias would have been around -0.06 (i.e., the true parameter 1.2), and that the bias is more sensitive to misclassification of exposure (i.e., nonsmoking spouses being truly smokers). If classification of subjects in this study had been 5 percent imperfect by the four parameters, the corrected RR would have been 1.2. 8.4 Confounding The decline of smoking in the US since the late 60's is reflected in the age distribution of either spousal ETS (particularly the prevalence of current smoking spouses), and self-reported ETS. Younger nonsmoking study subjects were more likely to have any ETS exposure than older persons in the analytic cohort of self-reported ETS. Since lung cancer rates increase exponentially with age, confounding by this variable occurred in the analyses of self-reported ETS. Data-based confounding (i.e., change in estimate) by SES and gender was not 0 O3 0~ 0 O~ I I I I I I I ,I I I I I

I I I I- I I I I I I I I I I I I I I 139 detected across the di~'~'erent an',dyses conducted in d~is ~tudy. However, estimates 'presented in this report are simultaneously adjusted for age, gender, 'race' and schooling as a proxy of SES. Uncontrolled confounding by age is unlikely to have occurred given the fact that in most analyses we grouped age by quinquennia, thus allowing variation within age levels. Inclusion of other variables in the Cox regression models did not affect the adjusted results reported in the previous chapter. No evidence of confounding by other risk factors such as the intake of foods containing carotcnoids, dietary fat, marital status, or history of ehronie lung disease, was found. 8.5 ETS-CH.D association is unexplained by miselassification of smoking status At the core of the Mantel-Lee bias argument against the scientific case for an ETS lung cancer and cardiovascular diseases association, is the contention that a very strong association of active smoking with those outcomes would be reflected by the misclassification of some smokers, more likely former smokers, among those selected into a study population .of nonsmokers. As shown in the NCR report and reproduced in an illustrative example above, the Mantel-Lee argument may be reasonable when discussing the ETS and lung cancer association, given the fact that the size of the effect (i.e., odds ratio) of active smoking on lung cancer risk is considerably large (22 among men in CPS if). This argument, however, cannot explain the association of CHD and ETS, as shown below. A review of the evidence from major cohort studies on active smoking and coronary heart mortality provides estimates of the RR that range from 1.58 to. 2.55 for current cigarette smokers (Fielding 1992). To set limits to the possible effect of the Mantel-Lee bias, we will follow the same approach illustrated in figure 3, but for the case of the ETS-coronary heart disease association. The same simplifying assumptions are used to provide the 0 0~

t40 figures: a closed coiaor~ of i miiiion subjects with aii observations except deaths from CHD, are censored at the end of the six-year follow-up, and deaths occurred at the mid-period. We assumed also a 3 percent misclassification of active smokers (Lee 1988). In addition, we assumed a CHI) mortality rate of 4.8 per 1000 subjects, with a 20 percent prevalence of current smokers, and a two-fold increased CI-ID death rate among smokers. It was found that misclassificafion of smoking status would not have a meaningful effect on the estimates of a such study (i.e., biasing the study from 1.0 to 1.03). The hypothesis of bias from misclassificafion of active smoking to explain the ETS lung cancer hypothesis as set forth by Mantel and Lee (Lee 1985) necessarily implies it should also explain the ETS-CHI) association. The number of reports on ETS and CHD has increased since this argument was fast presented in 1985 (Steenland 1992), and by refuting this statement, these studies further reduce the credibility of the argument of bias by misclassification of active smoking to explain the observed effects of ETS on lung cancer risk or any other major ETS- related disease. There is another major weakness of the Mantel-Lee or "active smoking misclassification bias", namely that most misclassified smokers are actually former smokers. The CI-[D-active-smoking relationship holds for current smokers and the increased risk is reduced by more than half by the end of the first year of cessation. Also, the risk of former smokers slowly approaches the risk of never smokers (Fielding 1992, US DHHS 1982). Therefore, the net effect of the potential bias argued by Mantel and Lee is negligible on the observed relationship between CHI) and ETS. Steenland made this point in a review of the ETS -CH]) association, noting that the effect of such bias would be about 2% (Steenland 1992). 1",3 0 0~ I I I I I I I ,I I I I I I ,I I

I I I I- I I I I I I I I I I I I I I 8.6 Causal Inference The research hypotheses outlined in the Introduction, cannot be rejected or verified on the basis of the results of a single study. The results of this study seem to support the hypothesis of a weak association of cumulative exposure to ETS with the risk of lung cancer among nonsmokers, increasing such risk by 20-30 percent. The lack of statistical significance of the estimates of the effect should not be confused with a null effect, because statistical significance depends heavily on numbers, and it has been shown that our study had little power to detect a RR of 1.2. In addition, a misclassification bias towards the null is likely to have taken place in this study, in an amount enough to dilute the RR estimate from t.3 to 1.2 (specificity <90%). It is not in the strength of the association that the ETS-lung cancer hypothesis finds support, ' but in other major criteria for causal inference in epidemiology. The time order of the observed association is a particular advantage of this study: the assessment of ETS exposure preceded the ascertainment of deaths. This criterion for causal inference is assured by the prospective nature of the study design. As discussed before, this design prevents the occurrence of recall bias. Consistency is the persistence of an association upon repeated test, and has two domains: survivability and replication (Susser 1991). Survivability stresses the number and severity of tests. This study adds survivability to the ETS and lung cancer hypothesis in at least the following ways. First, this study, controls more rigorously for age by using proportional hazards modeling and thus "stratifying" more finely for age, and at the same time it adjusted for SES, and many other potential confounders. Second, this study avoided the potential of recall bias more likely to occur in case- control studies. Last, this study also provided estimates for two independent sources of assessing ETS exposure: self-reported ETS and exposure from spousal smoking status, and the smoking status of spouses was doubly checked. Regarding repl!cability, most epidemiologic studies of lung cancer and ETS have consisted of non-smoking lung cancer cases among wives according to the smoking of o 0",

142 their husbands. The summary estimate from 37 epidemiologic studies, including this one, contrasting the risk of lung cancer of women according to their ETS exposure on the basis of their husband smoking status is still 1.2 (95% CI= 1. I-1.3). This study is consistent with a weak effect of ETS on the risk of lung cancer among nonsmokers. Based on previous knowledge of the joint effects of asbestos and tobacco smoke upon the risk of lung cancer, this study is a confirmation of the prediction that tobacco smoke involuntarily inhaled by nonsmokers exposed to asbestos will increase the risk of lung cancer above that of those exposed to either asbestos or ETS alone. Although based on few numbers, this study found suggestive evidence that this synergism might occur for both active smoking and ETS. It is important to note, at this point, that smoking spouses of CPS II nonsmokers smoked less than their counterparts of other US nationwide studies (e.g., Fontham 199 I). The distribution of pack-years indicates that only I percent of the entire cohort was exposed to 80+ pack years. The distribution of pack-years of CPS II smoking spouses of nonsmokers is skewed to the left with respect with to the SEER based case-control study population. Thus, the overall effect of ETS on lung cancer risk among nonsmokers is likely to be somewhat small because of the low level of ETS exposure in the cohort. Finally, the findings of this study are plausible in terms of pre-existing knowledge about the carcinogenicity of tobacco smoke components, in vitro and in vivo models, as weLl as from epidemiologic studies of active smoking. The biological plausibility of the ETS-lung cancer association is also well founded and it is based upon the evidence of harmful effects of ETS constituents leading to: I) an increased incidence of lower respiratory tract infections, additional episodes of asthma in children, reduced lung function, increased prevalence of middle ear infections and symptoms of upper respiratory tract infection in children (EPA 1992); and 2i an increased risk of CHD in adults in a similar pattern followed by active smoking (Steenland t992). Also, asbestos fibers increase cell proliferation and the occurrence of tumors (Kalburn 1992). Thus, this property of asbestos fibers added to the 0~ D~ 0~ I I I I. C I I I I I I I

I I I I- I I I I I I I I I I i I I I L43 genotoxicity properties of tobacco smoke are beneath the observed synergism of those two environmental hazards. [n summary, the following scientific facts lend biologic plausibility to the conclusion in epidemiologic studies like this that ETS causes lung cancer: 1) tobacco smoke from active smoking causes lung cancer as shown in epidemiologic studies, genotoxicity and animal data, 2) the same carcinogens found in MS and some other carcinogens perhaps more likely to reach the peripheral parts of the lung are present in ETS, 3) the levels at which ETS is present are consistent with those at which a risk is expected, 4) ETS is absorbed by nonsmokers in amounts at which a risk would be predicted, and 5) that the collective findings of epidemiologic studies like this one, strongly support a cause-effect relationship. 8.7 Conclusions 1. With respect to our first hypothesis, our study found that non-smokers exposed to ETS from current spousal smoking are at higher risk of fatal lung cancer than are non- smokers not exposed to ETS. However, we failed to provide precise estimates, and the 95% CI included the null value. Current spousal smoking increased the risk of lung cancer of non-smokers (both men and women) by 30% (0.8-1.9). Our study did not find an overall association with self-reported ETS exposiare. However, we found indication that missing data on reported hours of exposure to ETS may have introduced misclassification, thus biasing the results towards the null. 2. Our study found a weak dose-response relationship with pack-years of cigarettes smoked during marriage by husbands of nonsmoking women, but also this relationship was not statistically signficant (test for trend p---0.14). This relationship was not found for nonsmoking men. There was an 50 percent increased risk of of lung cancer among nonsmoking wives married to cigarette smoking husbands who smoked heavily during their marriage (36+ pack- years) [RR=I.5 (95% CI=0.8-2.8)]; these women represent the upper 17 percent of those married to ever cigarette smokers. l 0 O~ 0:~ I

t44 3. Consistent with the summary estimate of t.2 (95%CI=1.1-1.3) for the risk of nonsmoking women married to ever smokers from 36 published epidemiologic studies reviewed for this paper, this study reports an estimate of 1.3 for the risk of lung cancer among nonsmoking women married to current smokers (any type) (95% CI=0.8-1.9). 4. This study among nonsmoking CPS II participants suggests that there are greater than expected joint effects of ETS and occupational exposure to asbestos (p=0.18). If this relationship exists, it would resemble the known synergism between active smoking and asbestos. 5. The nonstatistically slgnificant association between ETS exposure from spousal smoking and the risk of lung cancer remained unchanged after adjustment for relevant potential confounders, and is not attributable entirely to misclassification of smoking status (i.e, misclassified smokers are included in a study restricted to nonsmokers). 6. Consistent with larger studies, a small validation study found that diagnosis of lung cancer from death certificates correctly classifies lung cancer deaths. Therefore, epidemiologic studies of lung cancer which rely on diagnosis from death certificates may still yield valid estimates of effect. I I I I I I I I I I I I i

145 I I I I I I I I I I I I I I References ACS: Cancer facts and figures- 1993. Atlanta: ACS, 1993. Akiba S, Kato H, Blot WJ: Passive smoking and lung cancer among Japanese women. Can Res 1986; 46: 4804-4807. Alavanja MCR, Brownson RC, Boice JD, and Hock E: Preexisting lung disease and lung cancer among nonsmoking women. AJE 1992; 136 (6): 623-632. Alavanja MCR, Brown CC, Swanson C, and Brownson RC: Saturated fat intake and lung cancer risk among nonsmoking women in Missouri. JNCI 1993 85: 23: 1906- 1916. Appel BR, Guirguis G, Kim I, Garbin O, Fracchia M, Flessel CP, Kizer KW, Book SA, and Warriner TE: Benzene, benzo[~]pyerne, and lead in smoke of tobacco product other than cigarettes. AJPH 1990; 80: 560-564. Auerbach O, Hammond EC, Kirman D, and Garfinkel L: Effect of cigarette smoking on dogs. II Pulmonary Neoplasms. Arch Environ Health 1970; 21: 754-768. Becher H, Zatonski W, and Jrckel KH: Passive smoking in Germany and Poland: Comparison of exposure levels, sources of exposure, validity and perception. Epidemiolo~ 1992; 3:509-514. Blot WJ and Fraumeni JF: Passive smoking and lung cancer. JNCI 1986; 77: 993- I000. Blot WJ: Lung cancer and occupational exposures. In Mizell M, and Correa P: Lung cancer: causes and prevention. Deerfield Beach: Verlag Chemie International, 1984. pp. 47-64. Boffetta P, Stellman SD, Garfinkel L: Diesel exhaust exposure and mortality among males in the American Cancer Society Prospective Study. Am J Ind Med 1988; 14: 403-415. Boring CC, Squires TS, Tong T: Cancer statistics, 1993. Atlanta: ACS, 1993. Breslow NE and Day N'E: Statistical methods in cancer research. Vol II. The design and analysis of cohort studies..World Health Organization, I.ARC pub. # 82, Lyon. 1987. Brilliant LB, Lepkowski/M and Musch DC: Reliability of ophthalmic diagnoses in an epidemiologic survey. AJE 1983; 118: 265-279. 0 O3 0

146 Brownson RC, Reif JS, Keefe TJ, Ferguson SW, and Pritzl JA: Risk factors for adenocarcinoma of the tung. AJE 1987; 125 (t): 25-34. Brownson RC, Alavanja MC, Hock ET, Loy TS: Passive smoking and lung cancer in nonsmoking women. ALPHA 1992; 82: 1525-1530. Butler TL: The relationship of passive smoking to various health outcomes among Seven-Day Adventists in California. Dissertation. Los Angeles: rdCLA. 1988. Butler W.h Lung cancer, spousal smoking status, and confounding AJE 1990; 131: 724. Calle EE and Terrell DD: Utility of the NAtional Death Index for ascertainment of mortality among Cancer Prevention Study II participants. AJE 1993; ,137 (2): 235- 241. Claxton LD, Morin RS, Hughes TJ, Lewtas J: A genotoxic assessment of environmental tobacco smoke using bacterial bioassays. Mutation Res 1989; 222:81- 99. Caporaso NE, Tucker MA, Hoover RN, Hayes RB, Pickle LW, Isaaq HI, Muschick GM, Green-Gallo M, Buivys D, Aisner S, REsau JH, Trump BF, Tollerud D, Weston A and Harris CC: Lung cancer and the debrisoquine metabolic phenotype. J-NCI 1990; 82: 1264-1272. Chilmonczyk BA, Knight GJ, Palomaki GE, Pulkkinen A J, Williams ,1 and Hadow JE: Environmental Tobacco Smoke exposure during infancy. AJ-PH 1990; 80 (10): 12O5-1208. Coleman P: Directory of on-going research on cancer epidemiology.. Lyon: IARC, 1992. Comstock GW, Bush TL, Helzlsouer K: Serum retinol, beta-carotene, vitamin E, and selenium as related to subsequent cancer of specific sites. AJE 1992; I35 (2): 115- t2t. Cox DR: Regression models and life tables (with discussion). J Royal Statistical Society. Series B 1972; 34: 187-220. Cummings KM, Markello S]', Mahoney MC, Marshall JR.: Measurement of lifetime exposure to passive smoke. AJE 1989; 130 (1): 122-132. Cummings KM, Zaig A, Markello S: Variation in sensitivity to environmental tobacco smoke among ~.dult non-smokers. IJE 1991; 20:121-125. I I I I I I I ,I I ,I

I I I I- I I I I I I I I I I I I I I I47 Dalager NA, Pickle LW, Mason TJ, Correa P, Fontham ET, Stemhagen A, Buffler PA, Ziegler RG. and Fraumeni IF Jr: The relation of passive smoking and lung cancer. Can Res 1986; 46:4808-4811. Devesa SS, Blot WJ. Fraumeni JF: Declining lung cancer rates among young men and women in the United States: A cohort analysis. JNCI 1989; 81: 1568-1571. Doll R, Hill AB and Kreyberg L: The significance of cell type in relation to the aetiology of lung cancer. Br J Cancer i957; I 1: 43-48. Doll R, Hill AB: Smoking and carcinoma of the lung. BMJ 1950; 2: 739-748. Emmons KM, Abrams DB, Marshall RJ. Ethzel RA, Novotny TE, Marcus BH, and Kane ME: Exposure to Environmental Tobacco Smoking in naturalistic settings. AJPH 1992; 82 (1): 24-28. Environmental Protection Agency Report: Respiratory Health Effects of passive smoking: Lung Cancer and Other Disorders. Washington: EPA. December 1992. Fielding JE: Smoking: Health effects and control. In Last JM and Wallace RB (eds): Maxcy-Roseanu-Last Public Health & Preventive Medicine. 13th ed. Norwalk: Appleton & Lange. 1992, pp.715-740. Fleiss JL and Gross AJ: Meta-analysis in epidemiology, with special reference to studies of the association between exposure to environmental tobacco smoke and lung- cancer: a critique. J Clin Epidemiol 1991; 44 (2): 127-139. Fleiss JL: Statistical methods for rates and proportions. 2nd ed. N. York 1981. Fontham ET, Correa O, Wu-Williams A, Reynolds P, Greenberg RS, Buffler PA, Chert VW, Boyd P, Alterman T, Austin DF, Lift J, and Greenberg SD: Lung cancer in nonsmoking women: a multicenter case-control study. Cancer epidemiology, biomarkers and prevention 1991; 1: 35-43. Fontham ET: Protective dietary factors and lung cancer. IJE 1990; 19 (3) Supple I, p. $32-$42. Friedman GD, Pettiti DB, Bawol RD: Prevalence and correlates of passive smoking. AJPH 1983; 73 (4): 401-405. Gann P, Coghlin-Strom J, and Hammond K: Measurement of bias due to misclassification in epiderniologic studies of lung cancer. AYE 1988; 128 (4): 920-921. Gao YT, Blot WJ, Zheng W, Ershow AG, Hsu CW, Levin LI, Zhang R, and Fraumeni J-F: Lang cancer among Chinese women. IJC 1987; 40: 604-609.

148 Garfinkel L: Time trends in lung cancer mortality among non-smokers and a note on passive smoking. JNCI 1981; 66 (6): I061-1066. Garfinkel L, Auerbach O, and Joubert L: Involuntary smoking and lung cancer: a case- control study. JNCI 1985; 75: 463-469. Garfinkel L, Boffetta P: Smoking and estrogen-related cancer. Data from the ACS studies. In Wald N and Baron I: Smoking and hormone related isorders, Oxford: Oxford Medical Pub, 1990. Garfinkel L, Boffetta P: Association between smoking and leukemia in two ACS prospective studies. Cancer 1990; 65: 2356-2360. Garfinkel L and Silverberg E: Lung cancer and smoking trends in the US over the past 25 years. CA 1991; 41: 137-145. Glantz SA and Parrnley WW: Passive smoking and heart disease. Circulation 199 l; 83 (1): 1-12. Glantz SA, Parmley WW: Passive smoking causes heart disease and lung cancer. J Clin Epidemiol 1992; 45 (8): 815-819. Glass A: Declinig rates of lung cancer in the US in young men and women. JNCI 199I; 83 (5): 368-369. Greenland S and Robins J: Conceptual problems in the definition and interpretation of attributable fractions. AJE 1988; 128 (6): 1185-1197. Greenland S: Modeling and variable selection in epidemiologic analysis. AJ-PH 1989; 79-(3): 340-349. Guess HA and Thomas E J: A rapidly converging algorithm for exact binomial confidence intervals around the relative risk in follow-up studies with stratified incidence-density data. Epidemiology 1990; 1 (I): 75-77. Hammond EC and Horn D: Smoking and death rates -report on forty-four months of follow-up of I87,783 men. JAMA 1958; 166:1 I59-I 172. Hammond EC: Smoking in relation to the death rates of one million men and women. I: Haenszel W, ed.: Epidemiological approaches to the-study of cancer and other chronic diseases. NCI Monogr 16: 127-204, 1966. Heath CW Jr.: Passive smoking: Enivronmental tobacco smoke and lung cancer. Lancet 1993; 341: 701. Hennekens CH and Buring JE: Epidemiology in Medicine. Boston: Little and Brown, t987. p. 89. I I 1 I I I I I I 0 0 O~ I I

149 I I" I I I I I I I I I I I I I I Hinds ~V, KoLonel LN, Lee J: Application of a job-exposure matrix to a case-control study of lung cancer. JNCI 1985; 75: 193-197.. Hirayama T: Non-smoking wives of heavy smokers have a higher incidence of lung cancer: a study from Japan. BMJ 1981; 282: 183-185. Hirayama T: Cancer mortality in nonsmoking women with smoking husbands on a large-scale cohort study in Japan. Prev Med 1984; I3: 680-690. Hole D.I', Gillis CR, Chopra C, Hawthorne VM: Passive smoking and cardiorespiratory health in a general population in the West of Scotland. BMJ I989; 299: 423--427. Humble CG, Samet JM, Pathak DR: Marriage to a smoker and lung cancer risk. AJPH 1987; 77: 598-602. Inoue R, Hirayama T: Passive smoking and lung cancer in women. Smoking and Health. Elsevier. L988, pp. 283-285. Janerich DT, Thompson WD, Varela L, Greenwald P, Chorost S, Tucci C, Zaman M_B, Melamed MR, Kiely M, McKneally MF: Lung cancer and exposure to tobacco smoke in the household. NE.IM 1990; 323: 632-636. Kabat GC and Wynder EL: Lung cancer in nonsmokers. Cancer 1984; 53: 1214-1221. Kalandidi A, Katsouyanni K, Voropoulu N, Bastas G, Saracci R, Trichopoulos D: Passive smoking and diet in the etiology of lung cancer among non-smokers. Cancer causes and control 1990; l: 15-21. Kellerman G, Shaw CR, Luyten-Kellerman M: Aryl hydrocarboxilase inducibility and bronchogenic carcinoma. NFAM 1973; 289: 934-937. Kelsey JL, Dwyer T, Holford R, Bracken MB: Maternal smoking and congenital malformations: An epidemiological study. J Epidemiol Community Health 1978; 32: 102-107. Kleinbaum D, et al : Epidemiologic Research (principles and quantitative methods). Lifetime Learning Publications, Belmont, 1982. Kilburn KH: Asbestos and other fibers. In Last JM and Wallace RB (eds): Maxcy- Roseanu-Last Public Health & Preventive Medicine. 13th ed. Norwalk: Appleton & Lange. 1992, p. 349. Kreyberg L: Histological lung cancer types. A morphological and biological correlation. Acta Pathol Microbiol Scand 1962; 157: 1-93.

150 Lam TH, Kung ITM, Wong CM, Lam WK. Kleevens JWL, Saw D, Hsu C, Sneviratne H, Lam SY, Lo KK, and Chart WC: Smoking, passive smoking and histological types in lung cancer in Hong Kong Chinese women. Br J Cancer 1987; 56: 673-678. Landis JR and Koch GG: The measurement of observer agreement for categorical data. Biometrics 1977; 33:671-679. Last JM: Dictionary. of Epidemiology. 2rid edition. IEA. New York: Oxford University Press, 1988, pp. 9-10. Lee PN in Lehnert G, Garfinkel L, Hirayama T, Schm~l D, Oberla K, Wynder EL, and Lee PN: Roundtable discussion of the Symposium "Medical Perspectives on Passive Smoking". Preventive Medicine 1984; 13: 730-746. Lee PN: Does breathing other people's tobacco smoke cause lung cancer? BMJ" 1986; 293: 1503-1504. Lee PN: Misclassification of smoking habits and passive smoking[ A review of the evidence. Int Arch Occup Env Hlth. Berlin: Springer-Verlag, 1988. Lee PN: Environmental Tobacco Smoke and Mortality. A detailed review of epidemJological evidence relating ETS to the risk of cancer, heart disease and other causes of death in adults who have never smoked. Basel: Karger, 1992. Lee PN: Environmental tobacco smoke and heart disease..I'AMA 1992; 267 (24): 3284. Lehnert G, Garfinkel L, Hirayama T, Schm~ihl D, [rberla K, Wynder EL, and Lee PN: Roundtable discussion of the Symposium "Medical Perspectives on Passive Smoking". Preventive Medicine 1984; i3: 730-746. Letzel H, Blumner E, Uberla K: Meta-analyses on passive smoking and lung cancer. Effects of study, selection and misclassification of exposure. Environ Technol Lett 1988; 9: 491. Levine M, Goldstein H, and Gerhardt PR: Cancer and tobacco smoking: a preliminary report. JAMA 1950; 143 (4): 336-338. Little RJA and Rubin DB: Statistical analysis with missing data. New York: Wiley, 1987. L6froth G: Environmental tobacco smoke: overview of chemical composition and genotoxic components. Mutation Res 1989; 222: 73-80. I I I I I I C. I I I ! I I I I

I I I I I I I I I I I I I I I I I I Lychou C: Passiv Rokning-Glom into Pipoch Cigarrokharna! [Passive Smoking- Don't forget Pipe and Cigar Smokers!] Lakartidningen 1986; 83: (23-27): 2373. Mantel N and Haenszel W: Statistical aspects of the analysis of data from retrospective studies of disease. Journal of the National Cancer Institute 1959; 22 (4): 719-747. Mantel N: Epidemiologic investigations: Care in conduct, care in analyses and care in reporting. J Canc Res Clin Oncol 1983; 105:113-I 16. Mantel N: What is the epidemiologic evidence for a passive smoking-lung cancer association? In: Kasuga H, Ed: ~door Air Quality. Berlin: Springer-Verlag, 1990: 341-347. Mantel N: Dubious evidence of heart and cancer deaths due to passi,,e smoking. J Clin Epidemiol I992; 44 (8): 809-813. Miettinen OS and Cook EF: Confounding: essence and detection. AYE 1981; 114 (4): 593-603. Miettinen OS: Theoretical Epidemiology. Principles of occurrence research in Medicine. New York: Wiley, 1985. Minna JD, Pass I-I, Glatsein E, Ihde DC: Cancer of the lung. In DeVita VT, Hellman S, Rosenberg SA: Cancer: PrincipLes and Practices of Oncology. 3rd edition. Philadelphia: Lippincott, 1989, p. 602. MMWR: Environmental Tobacco SMoke in the workplace: Lung cancer and Other Health Effects. MM~VR. 1991; 40 (25): 430-431. MMW'R: Discomfort from Environmental Tobacco Smoke among employees at worksites with minimal smoking restrictions. MMWR 1992; 41 (20): 351-354. MMWR: Preliminary Data: Exposure of Persons Aged >4 years to tobacco smoke -US, 1988-1991. ~UvlWR 1993; 42 (2): 37-39. Mulvihill JJ and Bale AE: Ecogenetics of lung cancer: genetic susceptibility in the etiology of lung cancer. In Mizell M, and Correa P: Lung cancer: causes and prevention. Deerfield Beach: Verlag Chemic International, 1984. pp. 141-152. National Cancer Institute: Cancer Statistics Review 1973-1986. NIH Pub. 89-2789. Bethesda: NCI. May 1989. National Cancer Institute: Cancer Statistics Review 1973-1988. NII-I Pub. 89-2789. Bethesda: NCI. May 1991.

152 I National Center for Health Statistics: 1988 National Health Interview Survey- Occupational Health Supplement Data. WONDER Documentation on tapes number DSN=CC36.NHIS 88. OCHEALTH. National Research Council. Committee on Passive Smoking. Board on Environmental Studies and Toxicology: Environmental Tobacco Smoke. Measuring exposures and assessing health effects. Washington, DC: National Academy Press, 1986. O'Neill IK, Brunnemann KD, Dodet B, and Hoffmann (eds): Environmental carcinogenesis methods of analysis and exposure measurement. Vol 9 - Passive smoking. WHO/IARC, Lyon. New York: Oxford University Press. 1987. Pershagen G: Passiv Rokning-Glom into Pipoch Cigarrokharna! [Passive Smoking- Don't forget Pipe and Cigar Smokers!] Lakartidningen 1986; 83: (23-27) 2373. Percy C, Stanek E, and Gloeclder L: Accuracy of cancer deaths certificates and its effect on cancer mortality statistics. AJPH 198 I; 71 (3): 242-250. Peto R, Lopez AD, Boreham J, Thun M J, and Heath CW: Mortality from tobacco in developed countries: indirect estimation from national vital statistics. Lancet 1992; 339: I268-1278. Pron GE, Burch JD, Howe GR, Miller AB: The reliability of passive smoking histories reported in a case-control study of lung cancer. AJE 1988; 127: 267-273. Redmond DE Jr: Tobacco and cancer: The first clinical report, 1761. NEJM 1970; 282: 18-23. Reif JS, Dunn K, Oglivie GK, and Harris CK: Passive smoking and canine lung cancer risk. AJE 1992; 135: 234-9. Repace JL and Lowry AH: Indoor air pollution, tobacco smoke, and public health. Science 1980; 464-472. Repace JL and Lowry AH: Tobacco smoke, ventilation, and indoor air quality. ASHR.AE Trans 1982; 88: 894-914. Robins J: Risk assessment - Exposure to environmental tobacco smoke and lung cancer. In National Research Council. Committee on Passive Smoking. Board on Environmental Studies and Toxicology: Environmental Tobacco Smoke. Measuring exposures and assessing health effects. Washington, DC: National Academy Press, 1986., pp. 294-337. Rothman KJ and Boice JD: Epidemiologic analysis with a programmable calculator. 2nd ed. Brookline: Epidemiology Resources Inc, 1982. 0 0"~ 0 0 I I I I I I I I I I I

I I I I- I I I I I I I I I I I I I I I Rothman Y-d: Modern epidemiology. Boston: Little, Brown and Company. pp. 165- 166. SAS Institute: SAS Technical report P-217". The PHREG Procedure. Cary: SAS hastitute Inc, 1991. Selikoff H, Hammond EC, Churg ]': Asbestos exposure, smoking and neoplasia. .lAMA 1968; 204: (2): 104-110. Shephard R J: The risks of passive smoking. New York, Oxford University Press, 1982.p. 33. Shimkin MB: Some Historical Landmarks in Cancer Epidemiology. InSchottenfeld D: Cancer epiderniology and prevention. Springfield: Thomas, I975. pp. 62-63. Shopland RD, Eyre HI, and Pechacek TF: Smoking-attributable cancer mortality in 199 I: Is lung cancer now the leading cause of death among smokers in the US? JNCI 1991; 83 (16): 1142-1148. Sobin LH, Yesner R: WHO International histological classification of tumors, l- Histological typing of lung tumors. 2nd. edition. Geneva: WI-IO, 1981. Spitzer WO, Lawrence V, Dales R, Hill G, Archer MC, Clark P, Abenhaim L, Hardy J, Sampalis J, Pinfold SP, and Morgan PP: Links between passive smoking and disease: a best evidence synthesis. A report of~the Working Group on Passive Smoking. Clin Inv Med 1990; 13 (i): 17-42. Steenland K: Passive smoking and the risk of heart disease..IA/vIA 1992:267 (i): 94- 99. Stellman SD, and Garfinkel L: Smoking habits and tar levels in a new American Cancer Society prospective study of 1.2 million men and women..INCI 1986; 76: 1057-1063. Stellman SD, Boffetta P, Garfinkel L: Smoking habits of 800,000 American men and women in relation to their occupation. AJIM 1988; 13: 43-58. Stellman SD: Sweetener usage in America. A brief history and current usage patterns. In Williams GM: Sweeteners: Fiealth Effects. Princenton: Princenton Scien Pub 1988, pp. 1-18. Stockwell FIG, '9oldman AL, Lyman GH, Noss CI, Armstrong AW, Pinkham PA, Candelora EC, and Brusa MR: Environmental tobacco smoke and lung cancer risk in nonsmoking women. JNCI 1992; 84: 1417-1422. 0 C~ C~ 0 ",0

i54 Stolwijk JAJ and Kessler II: Workgroup report. Assessing low-risk agents for tung cancer: methodological aspects. IJE 1990; 19 (3): $84-$86. Susser M: What is a cause and how do we know one? A grammar for pragmatic epidemiology. AJE 1991; I33 (7): 635-648. Thomas DB: Cancer. In Last JM and Wallace RB (eds): Maxcy-Roseanu-Last Public Health & Preventive Medicine. 13th ed. Norwalk: Appleton & Lange. 1992, pp. 811- 826. Thun MJ, Namboodiri MM, Heath CW Jr: Aspirin use and reduced risk of fatal colon cancer. NEJM 1991; 325 (23): 1593-1596. Thun M.I', Calle EE, Namboodiri M.M, Flanders DW, Coates R J, Byers T, Boffetta P, Garfinkel L, Heath CW: Risk factors for fatal colon cancer in a large prospective study. JNCI 1992; 84 (19): 1491-1500. Thun MJ, Altekruse SF, Namboodiri MM, Calle EE, Myers DG, and Heath CW: Hair dye use and risk of fatal cancer in US women. JNCI 1994; 86 (3): 210-215. Trichopoulos D, Kalandidi A, Sparros L, MacMahon B: Lung cancer and passive smoking. Int J Cancer 1981; 27: 1-4. Trichopoulos D, Molio F, Tomatis L, Agapitos E, Delsedime L, Zavitsanos X, Kalandidi A, Katsouyammi K, Riboli E, Saracci R: Active and passive smoking and pathological indicators of lung cancer risk in an autopsy study. JA.MA 1992; 268 (13): 1697-1701. US DHHS: The Health consequences of involuntary smoking: a report of the Surgeon General. Washington DC: US GPO, 1986. US DHHS: The health consequences of Smoking: Cancer. A report of the Surgeon General. Washington, DC: US GPO, 1989. Vandenbroucke JP: Passive smoking and lung cancer: a publication bias? BMJ 1988; 296:391-392. Vincent RG: Lung cancer: A twentieth-century disease. In Mountain CF and Carr DT: Lung cancer. Current status and prospects for the future. Austin: University of Texas Press, 1986. Wall MA, Johnson S, Peyton J, Neal LB: Cotinine inte serum, saliva and urine of nonsmokers and smokers. AJPH 1988; 78:699-701. Wells A J: An estimate of adut mortality in the US from passive smoking. Environ Int 1988; 14:249-265. I I I I I I I I I ,I I I I I I I

I I I I- I I I I I I I I I I I I I I I55 Wells AJ: Passive smoking and lung cancer: a publication bias? BMJ 1988: 296:1128. Weston A, Caporaso NE, Taghizadeh K, Hoover RN, Tannenbaum, Skipper PL, Resau JH, Trump B J, and Harris CC: Measurement of 4-aminobiphenyl-hemoglobin adducts in lung cancer cases and controls. Can Res 1991; 5 I; 5219-5223. Wetzel B and Anderson R: Tumor registries. In Lung Cancer. The evolution of concepts. Vol II. New York: Field & Wood, 1989. pp. 1-23. WHO: International Classification of Disease. 9th Revision. (2 Volumes). Geneva: WHO, 1979. Willett W: Nutritional epidemiology. New York: Oxford University Press, 1990, pp. 83-84. Wynder EL and Goodman MT: Smoking and lung cancer: some unresolved issues.Epidemiol Rev 1983; 5: 177-207. Wynder EL, Graham EA: Tobacco smoking as a possible etiologic factor in bronchogenic carcinoma. A study of six hundred and eighty-four proved cases. JAMA 1950; 143: 329-336. Wynder EL, Mabuchi K, Hoffmann D: Tobacco. In Schottenfeld D: Cancer epidemiology and prevention. Springfield: Thomas, 1975. p. 106. Yamagiwa K, Ichikawa K: Experimental study of the pathogenesis of carcinoma. J Cancer Res 1918; 3: 1-29. 0 0", 0 0", ',0

158 APPENDICES A References and Tables of published studies and Metanalysis of ETS-Lung Cancer B Abbreviations C CPS II Questionnaires and Instructions i ! I ! I I I I I I I I I I

Appendix A I Case-control study Authors: Trlchopoulos at Source: L~ncet. 1983 Country: Greece 2 Casa-conlrol sludy Authors: Chang &Fung Source: Grundmarm, t982 CounW:Hong Kong 3 Case-control sludy Authors: Corm,, et zl. Source: Lancet 1983 CounW: USA 4 Case-conlrol sludy Authors: Kaba! & Wynder Source: Cancer 1984 Country: USA 5 Case-control study Aulhors: Bultler at el. Source: Mlzell 1983 Counlry:USA Male-analysis 0l 36 studies on ETS-lung cancer ETS.Lung Cancer In CPS II Smoking habits of spouse CR Smoker Nonsmoker Total Cases 53 24 77 2.075 Controls 116 109 225 Total 169 133 302 In (OR) Vat In (OR)IN In(OR) WI'In(OR)I 0.73 0.07833 12.7666 9.31954 Smoking habits of spouse Smoker Nonsmoker Tolal Cases 34 50 84 0.752 Conlrols 66 73 139 Toter 100 123 223 Smoking habits of spouse Smoker NonSmoker Total Cases 1 7 14 3 I Conlrols 87 226 313 Total 104 240 344 Smoking habits of six)usa Smoker Nonsmoker Total Cases 18 t 8 36 Controls 20 17 37 Total 38 35 73 In (OR) Vat In (OR)IN In(OR) Wl'ln(OR)l -0.285 0.078262 12.7776 -3.639801 a:t In (OR) Vat In (OR}I/V In(OR) Wi'ln(OR)l 3.154 1.1486 0.146171 6.8413 7.859t84 ~ in (OR) Vat In (OR)I/V In(OR) Wl'ln(OR)l 0.85 -0.163 0.219935 4.54681 -0.738942 Smoking habits of spouse C~ Smoker Nonsmoker Total Ceres 38 14 52 0.814 Controls 220 66 286 Tolal 258 80 338 In (OR) Vat In (OR)t/V In(OR) Wl'ln(OR)i -0,205 0.117441 8.51489 -t.749333 Pago 1 ~690E~9890~,

9690~9~90~ Appe nd~x A Meta-analysls of 36 studies on ETS-lung cancer ETS-Lung Cance; In CPS II 6 Case-conlrol study Authors: Gatltnkal at al. So, u~ce: JNCI 1965 Country:USA 7 Case-control study Authors: Wu et Source: JNCI 1985 Count~/: USA 6 Case-control study Authors: Aktba et al. Source: Can Res 1988 Country: Japan 9 Case-control study Authors: DaMager et =d. Source: Can Res 1986 Country: USA tO Case-control study Authors: L~e a! al. Source: BJC 1966 Counm/: UK Smoking habits of spouse Smoker Nonsmoke¢ Total Ca~as 91 43 134 Controls 254 140 402 Total 345 191 536 Smoking habits of spouse Smoker Nonsmoker Total Cases ? ? 29 Controls ? ? 62 Total ? ? 91 Smoking hablts of spouse Smoker Nonsmoker Total Cases 76 3 7 t 13 Controls 197 163 360 Total 273 220 493 Smoking habits of spouse Smoker Nonsmoker Total Cases ? ? 48 Controls ? ? 466 Total ? ? 514 Smoking habits of spouse Smoker Nonsmoker Total Cases 30 1 7 4 7 Controls 59 37 96 Total 80 54 .143 CR In (OR) vat In (OR)IN In(OR) WI'In(OR)I 1.233 0.2095 0.044939 22.2526 4.6627498 C~ In (OR) Vat In (OR)IN In(OR) Wl°ln(OR)l 1.2 0.1823 0.140231 7.t311 1.3001532 (CI-0.6-2.5) CR in (OR) Vat In (OR)IN In(OR) WI°In(OR)i 1.908 0.6461 0.050726 19.7139 12.73713 C~ in (OR) Vat In (OR)IN In(OR) Wi'ln(OR)l 1.47 0.3853 0.11947 8.3703 3.2247627 (95% CI-0.76-2.63) (~ In (OR) Vat In (OR)IN In(OR) Wl'ln(On)l 1.107 0.1014 0.136133 7.34576 0.744~988 Page 2 B~III

m mm n m m mm m m m m m mm m m mm L690~9890~ Mall-analysis ol 36 sludle$ on ETS-lung cancer ETS-Lung C~,ncer In CPS II 11 Case.control study Authors: Gao at am. Source: IJC 1987 Country: China 12 Case-control study Authors: Brownson el el. Source: AJE 1987 Counvy: USA 13 Case-control study AuthOrs: Koo el el. Source: IJC 1967 Country: Hong Kong 14 Case-conb'ol study Authors: Pershagen et Source: AJE 1987 Country: Sweden 15 Case.control study Authors: Humble et el. Source: AJPH 1987 Counuy: USA Smoking habits ol spouse CEt Smoker Nonsmoker Total Cases 246 t 90 436 0.794 Controls 375 230 605 Total 621 420 1041 Smoking habtts ol spouse Smoker Nonsmoker Total CILSeS 4 1 5 1 9 Controls 6 4 t 4 7 Total 10 56 66 Smoking heb~.ts ol spouse Smoker Nonsmoker Tolal Cases 5 t 35 86 Controls 66 70 136 Total 117 105 222 Smoking habits of spouse Smoker Nonsmoker Tolal Cms 37 44 01 Conlrols 153 215 368 Total 190 259 449 In (OR) Vat In (OR)I/V In(OR) WI'In(OR)I -0.231 0.016343 61.1894 -14.10657 CR In (OR) Vat In (OR)I/V In(OR) Wl'ln(OR)i 1.022 0.6001 0.507724 1.96958 1.1818572 (]:] In (OR) Vat In (OR)I/V In(OR) Wl'ln(OR)l 1.545 0.4353 0.077617 12.8639 5.6085763 CR In (OR) V~ In (OR)IN In(OR) Wl'ln(OR)l 1.182 0.1669 0.060941 16.4092 2.7391618 Smoking hab~t~ ol spouse CR Smoker Non~T~oker Total Cases 20 8 29 3.203 Controls 128 164 292 Total 148 172 320 In (OR) Vat In (OR)I/V In(OR) Wl'ln(OR)i 1.1641 0.18091 5.29352 6.162334 m m m

Appendix A Meta,-analysle of 36 etud~es on ETS.bng cancer ETS-Lung Canc~' In CPS II 16 Smoking habits of spouse Case-control study Smoker Nonsmoker Total Authors: Lain et al, 1987 Cases 115 84 199 Source: BJC, 1987 Controls 152 183 335 Cou~V,/: Hong Kong Total 267 287 534 17 Smoking habits of spouse Case-control eludy Smoker Nonsmoker Tolal Author's: Lain & Chang Cases 37 23 60 Source: Smoking and Health t987Controls 64 80 t44 Cou~n/: Ho, ng Kong Total 101 103 204 1 8 Smoking haldls of spouse Case.control study Smoker Nonsmoker Total Aulhors: Shlmlzu. 1988 Cases 53 37 90 Source: Toh J Exp Mad 1988 Conlrols 9 ! 72 163 Country: Japan Tolal 144 109 253 19 Smoking hablls ol spouse Case-control sludy Smoker Nonsmoker Total Authors: Inoue 191~8 Cases 18 4 22 Source: Smoking and Hearlhot988Conlrols 30 1 7 47 Cell numbers from Lee, 1992 Total 48 21 69 20 Smoking habits of spouse Case-conl~'ol study Smokar Nonsmokar Total Authors: Gang, 1987 Cases 34 20 54 Source: Smoking and Health 1987Controls 41 52 93 Count~/: China Total 75 72 147 P~e4 CR in (OR) V~ in (OR)l/V In(OR) WI'In(OR)I 1.648 0.4997 0.032644 30.6336 15.308276 C~ in (OR) Vat in (OR)I/V In(OR) WI'In(OR)I 2.011 0.6986 0.09863 10.1389 7.0826849 C~ in (OR) Varin (OR]IN In(OR) Wl'ln(OR)l 1.133 0.1252 0.070773 14.1297 1.768766 CR in (OR) Vat in (OR)IN In(OR) Wi'ln(On)I 2.25 0.8109 0.343758 2.90903 2.3590176 C1-(0.91-7.1) CR in (OR) Vet In (OR)I/V In(OR) WI'In(OR)I 2.156 0.7683 0.123033 8.12792 6.2446766 '8690~9890~ m m

Appendix ^ 21 Case-control sludy Authors: Katada, 1988 Source: Gan No Rlnsho 1988 Count~/: 22 Case.conlrol study Aulhors: Svensson, 1989 Source: Acta Oncol 1989 Counuy: Sweden 23 Case.control study Authors: Sobue e,t el. 1990 Source: Gen No R~nsho 1990 Counlty: Japan 24 Case.control study Aulhors: Janedch el el. 1990 Source: IJE 1991 Ceils esl~maled horn EPA. 1992 CI from authors, p. 834 Count,n/: USA 25 Case-conlrol study Authors: Wu-Witllame 1990 Source: BJC1990 Country: China Male-analysis ol 36 studies on ETS-|ung cancer ETS-I.ung Cancer In CPS II Smoking h~bits of spouse CR Smoker Nonsmoker Total Cases 17.5 0.S 17 8.448 Conlrols 14.5 3.5 1 7 Tolal 32 .4 34 Smoking habits of spouse Smoker Nonsmoker Tolal Cases 24 10 34 Conlrols 114 60 174 Tolal 138 70 208 Smoking habits of spouse Smoker NonsmoP~r Total Cms 80 84 144 Controls 395 336 731 Total 475 400 875 Smoking habits of spouse Smoker Nonsmoker Total Cases 14 7 44 191 Controls 153 38 t 91 Tolal 300 82 382 In (OR} Va~ I~ (OR)t/V In(OR) Wl'ln(OR)l 2.134 2.411823 177.3 0.4026 1.263 0.2336 0.167105 0.41462 0.8847924 in (OR) Va~ in (OR)I/V In(OR) WI'In{OR}I 5.98425 1.3980101 CR in (OR) V~ in (OR)I/V ~(OR)WI'In(OR)I !.063 0.0614 0.033633 29.7326 1.8246736 UR0(~R 1.2773 tJ)wOR 0.8851 ~ in (OR) Vat in (OR)iN In(OR) WI'In(OR)I 0.93 -0.073 0.071378 14.0099 .I.016711 (Cl-0.55-1.57) Smoking h=t~tts o! spouse Smoker Nonsmoker Total Ccsas 205 212 417 Conlrols 331 271 602 Total 536 493 1019 C]~ ~ (OR) V= in (OR)IN in(OR)Wl*ln(OR)l 0.792 -0.234 0.016306 61.3263 -14.32435 Page 5 6690~:9890~

MeP-~nalysls of 38 studies on ETS-lung cancer Proposal ETS.Lung Cancer In CPS II 26 Case.conUol study Aulhors: Kid)at el id. 1990 Source: Toxicology Forum 1990 Counln/: USA Smoking habits o! spouse CR In (OR} Vat In (OR)I/V In(OR) Wl'ln(OR)l Smoker Nonsmoker Total Cases 48 41 89 1.026 0.0252 0.001825 16.1747 0.4076721 Controls 129 113 242 Tolal 177 154 331 27 Case.conl~ol study Authors: Kalandtdi et el. 1991 Source: EPA 1992, Lea t992 Country: Greeco Smokk~g habits o! spouse C]~ In (OR) Vat In (OR)INN In(OR) WI'Ir,(OR)I Smoker Nonsmoker Tolal. Cases 64 26 90 1.573 0.4528 0.089715 11.I465 5.0466929 Controls 72 46 118 Total 136 72 208 26 Smoking hal:dis o~ spouse Case-control study Smoker Nonsmoker Total Authors: Llu el al. 1991 Cases 45 9 54 Source: IJEE t991 Controls 176 26 202 CountJy: China Total 22 t 35 256 29 Smoking habits of spouse Case-conb, ol study Smoker Nonsmoker Total Authors: Fontham el el. 1991 Cases 294 126 420 Source: C~ul Epld Biota Prey 1991 Controls 492 288 780 Country: USA Tolal 786 414 1200 CR In (OR) Vat In (OR)IN In(OR) WI'!n(OR)I 0.739 -0.303 0,177477 5.63454 -1.706982 (~ kt (OR) Vat ~n (OR)t/V In(OR) Wl'ln(OR)l 1.366 0.3118 0.016843 59.3732 18.511359 30 Smoking habits of spouse Case-control study Smoker Nonsmoker Total Authors: Brownsonsl a1.1992 Cases 215 213 431 Source: AJPH 1992 Conlrols 698 568 ! t 66 Total 816 76! 1567 CR In (OR) Var in (OR)INN In(OR) Wi'ln(OR)l 0.972 -0.028 0.012715 78.6465 -2.223tt4 Page 6

m m m mm mm m m mm m m mm m m mm mm Apl~ndix A Mela-anslysls el 38 studies on ETS-lung cancer Proposal ET$-Lung Cancer In CPS II 3 t Smoking habits of spousa Case-control study Smoker Nonsmoker Total Aulhors: Slockwell0 1992 Cases ? ? 210 Source:JNCI, 1992 Controls ? ? 30 t Total ? ? 5! t CR in (OR) V=r I~ (OR}iNN In(OR) Wl'ln(OR|l 1.60 0.47 0.104496 9.56974 4.497914S (95% CI-0.8-3.0) Vat- -0.015 32 Smoking habits of spouse Case-control sludy Smoker Nonsmoker Tolal Authors: Llu, 1993 Cases 25 1 3 38 Source: AJE 1993 Conlrols 37 32 69 Tolal 62 45 107 CR in (OR) Var in (OR)t/V In(OR) Wl'ln(OR)l 1.663 0.5087 0.1752 5.70776 2.9037908 LIL 3.7778 LL 0.7322 Summary osllmetss ol 32 case-control studies Summm~ in OR (Precision-based}- 0.147 Summary OR (Precision-based),- 1,15g Summmry V~' (In OR) (Prectslon-b~sed)- 0.002 SummlM~t SD (in OR) (Pmclskm-ba~d}- 0.042 Total Total Total 5.990523 571.669 84.272461 Summary Lower 95% OR (Pmdslon-ba=ed)- 1.068 Summit/ Upper 95% OR (Pxeclslon-bazed)- 1,258 Page 7 T, OLOZgSg07--,

Appendix A 33 Cohort sludy Authors: Hlray~ma. 1981 Source: Lancet 1981 34 Cohort study Authors: Gadinket, 1981 Source: JNCI, 1983 35 Coho~t study Authors: Hole el al.o 1989 S~urce: BMJ 1989 36 Cohort Study Authors: Butler 1989 Source: Dissertation UCL~, 1988 Summary estlmalas of four cohort studtss Mats-analysis of 36 studies of~ ETa-lung cancer Proposal ETS-Lung Cancer In CPS II Cases Total Cases Total Ca~s Total Smoking habit~ of spouse ~2vtCl In(SMR)welghts Smoker Nonsmoker Tolal 142 32 174 1.5 0.4055 142 69645 21895 91540 Va~{SMR) 0.047 Smok~ h=bR= of spouse Smoker Nonsmoker Total ~ In(SMR)walghts 88 65 153 49487 127252 176739 1.17 0.157 88 Va~ {SMR) 0.018 Smoking habits of spouse Smoker Nonsmoker Total i:~ In(RR) weights 7 2 9 2.1 0,7419 7 1538 917 2455 Smoking habits of spouse FR Smoka~' Nonsmoker Tota, I In(RR) weights Cases 2 6 B 2.01 0.6981 2 Noncasee 3128 6071 9199 Total 3130 6077 9207 Ovara# In(RR)- 0.326285 Overall RR- 1.3858 Sum ol welghl=- 239 veralt V~'lance- 0.0042 Overatt SO- 0.0647 Summa~' Lower 95% RR (Preclskmobased)- 1.221 Summa~f Upper 95% RR (Prectsk:m-based}- 1.573 Tolal 239 WI'In(RR)i 57,576 0.0382923 W|°In(RR)I 13.8163 0,0267483 WI'In(RR|I 5.19356 0,6428571 WI'In(RR)I t.39827 0.6666667 Total 1.3745643 gOLOg9890Z ll I

~ ~ m m m m m mm m m m m Appendix A Meta-an~,lysls ol 36 f, tud~ef, on F.TS-$ung cancel P;oposal ETS-Lung Cancer in CPS II Summary over 36 studies Ovara, In (RR) o! 36 etudes. 0.200149 Ovarlll (RR) o~ 36 slud~as- 1.221585 S.mmary V~r (in RR) (Preclskm.based). 0.00~234 Summary SD (in RR) (Precision-based). 0.035122 Summa~f Lower 95% RR (Preclslon-based}o 1.14032! Summ~y Upper 95% RR (Precision-based)- 1.30564 Page 9 EOLOZgEgo~

Appendix B Abbreviations AC5 CHD CP5 ! CP$ I~ I~PA [ARC ~CD-9 NCHS NCI NRC SEER S$ US DHHS WHO An~rk~n Cmcc~ Soc/cty Coronary hem disca~ ~r ~ven~on S~y ~o~ A~ for R~h ~o~ ~on ofDi~ ~ ~cn (1~5) Mo~i~ ~d Mo~ W~y ~v~w M~ (~) N~ ~ for H~ Na~on~ ~ Nafio~ ~ ~ Nu~fion Nafio~ ~h ~ S~e~ ~i~olo~ S~no~ s~ Si~ to~ ~ ~ N~on~ ~S~ey 0 0~ ~0 0 0 i I I i I I I ,I I I I I

I I I I I I I I I I I I CANCER PREVENTION STUDY II ~ Group , QUESTIONNAIRE FOR MEN "'-. ~ .J Resea~c~No." l~No. Person No. I. Name; 2. Date o~ birth: Month~__ Year 3. Howold am you now? 4. Cummt weight wi~ indomdo~ng: ,, 5. Weight 1 year ~go~" 6. Height (without shoes): ft. , FAMILY HISTORY (IN RELATION TO CANCER): I. F~I ~ ~ Ib~owing treble ~ completely a~ pos~ for parents, Ixo<heP+ a.~d dste~. UST ONE B~ IF IF 0~, OlD ~ RE~ PER ~E: ~ G~ ~E P~ ~ IF ~," (Ck~e ~ ~ Stst~) THIS Pl~qSON? GIVE AT HAVE ~F_.R? (C,V~ O~) AGE OEATH (O¢~e 0~) SPECIFY WHAT TYPE OF CANCER AGE? 2. When you were born, a) Howoldwas your motheC? b) How old wasyour fatty'? HISTORY' OF DISEASES: b) Date o~ first 2. Place a ¢hec~.~al~ by the fo#owing dlse~ses o¢ If "yes," spec~ type and date(~) c~ operation(s): 6. How n~ny times have you had c~Ids o~ flu in the past twelw mo~ths? , O~ 0 0

DIET: 1. On the average, how many days per week do you eat the following foods? (if less than Once a week, ~ at lea.st twice a mo~tl'~ write II2-) Brussels spmuts...~ Chocolate __ 2. How many days a week do,you eat ~e fo(Iow~ng fried foods? Fded eggs,, Fr~ed haml~Jrgers Fried bacon or beef. Fded chicker~sh Other fried foods French fries~ DO NOT EAT FRIED F~.OS t-'] 3. Do you eat a vegetaxian d~et? I-] Yes ~-I No ff "ye.s," what bJpe at~d for how many yearz? ......... Has nero beec a major change in your diet in last lOysam? I-1Ye~ [] No If "yes," what was d~e charge? 5. ~) Do you now or h~we you eve~- added ~ sweetenerz (saccharin or cycdamates) to coffee. tea, or other ddnks or foed? b) If ever used artificial sweeteners, indicate amount pet day and for howlong. Packets: No. pc," day Yezr~ Orops: No. per day Yeats Ta~ets: No. perday ,, Yeats 6. Do you get your ddnldnc.~ water fmm:l-] Citysupf~y [] Palate wetl [] Other (spec~), 7. DO you add a~/substances to soften your drinldng water? [] Yes [] No 8. How many cups, glasses, or ddnk~ Of thes~ bever- :~ges do YO~ usually ddnk a day, and for how many yem's? (If yo,J no Io~je¢ drink :z listed beverage, or ~0ur pattern fl~ ch~ed in ~he ~ ten years, indi- cate pm,Ao~s and cu~ent amourC¢ If less than onoe a dWo but at least three tJn~s a week, write 1/2.) Whole milk (n<X sldm n~lk] 'Caffeinated coffee !Oecaffeinated coffee "re= Diet soda or diet iced tea HaKI Ik~uor MEDICATIONS AND VITAMINS: 1. How many times in the last mo~th have ~K~J used the fol~ a.,',d how Iortg have y~0~ used them? (If no~e, write 0; If used only occa.sior~ly, wrife V'damin A V'darnin E Mult~V'~rntns BkxKI Pressure Oiuretk~ (wate¢ p~lls) Heart VaJium Ubdum Prescription sleepir~ p~Ls O~ 0 0 O~ I I I i I I i ! I i I

I I I I I I I I I I I CURRENT PHYSICAL CONDITION:1. H~w much exemise do you get (wo~ ot piay)? [] None [] Si~ht [] Moderate [] Heavy 2. On the average, how many hou~ (~0 you ~eep each night? . 3. On the average, ~ many t~'nes = month do you have klsomni=? .[] None 4. Within the last month, have you no(P..ed: a)PalnfulorfrequentudnatJon? []Yes iT) N0 b) An unusual discharge from you¢ penis? 5. Do you noLice pa~n~ in your legs ~an you w.~dk which go a~ay ~e~ you ~e~? [] Yes [] No If "yes," how many yea~ ~ you had these pak~s? 6oAmyous~c~<attheWesanttJme? C]Yes []No If "yes," w~th whai disecse or cond~n? HABITS: 1. Whether or not you ,smoke, on the average, how many hour~ • day am you exposed ~ c~Jarette smoke of o~: At home ~ At work ,. 2. Do you now or have you ever smoked dgamttes. dgars or pipes, at least one a dcf fo~ one Lime? ~ Yes [] No if n~ver smoked, skip to questk~ 8. If you currently smoke c~gsmttes" dg~rs o~ p~pes" fill in the ~focm~Lioa below:. Average number ~a t~-gan smokin9 :INHALATION: loft=de moderately' Inhale dee~ Total years of smokin9 Ye~rs smoked filtered cigarettes Years smoked non*filtered dgamttes 4. Currant brand of dgamtts: b) [] Non-filter [] Rlter [] Mentt~ c) Years smoked this brand:. If, ~ou have quit smoking dgamttes" dgars or ~pes" fill in the intormalJon below:. i Average number smoked ~ day :Age began smok!r INHALATION: In~aled n~demzel), Inhaled deep~ Total yea~ smok~. , Yea~s smoked fiZter~,, dgam~e= Yearn smoked non-filtered dgerettes 6. Last brand of dgarette smoked: ....... a) Size:[] Regumr [] K~ng[] 100ram b) [] Nan-fitter[] Rlter [] Men~'x~ c) Yea~s smoked this brand: 7. Current and •x-cigarette smoker& fill in the following infom~t~n fO~. 1) The first brand smoked ~ulady;, and 2) The brand of cigarette ~rnoked for the longest pednd of time. 1. 1 2. . 8. Have you ever eta•wed robs:co at least once a week fo~ =t least one ye~? I"1 Yea [] No If "no," skip to question 9. a) Age began cttew~ng tobacco: b) How many times a week? c) For how many ye~s? d) DO yo~ sLill chew to03cco? ' [] Yes [] No 9. H~ve you ever used snuff at least orce= week for al leas= one year?. C] Yes I"1 No If'no," skip to "OieL" " • ) ~ began using snu~ b) How many Limes a week? c) For how many yeats? 4) Oo you sLi|! use snuff? ~ Yes [] No 0 0 0

OCCUPATIONS: 1. Wh~ is your current occup~dan ~ wh~t ~ ~r ~es? How m~, yea~: 2. Ifretimd, wt~tw~syouclastoccupwj~?.. 3. What olhe~job have you heid fo~ ~e longest pedod of time?. How many ye~: 4. W~t ~meof daydo you Oo yau wo~ r~ng st~? 5. How many h~Jts a week do you v,~or~ on: paid jobs . ~lu~tee4" ¢ In yo~" wock or dailylite, a~(wera) you regul~ly Coai Tax/PitcW~ Diesel F_~ne F.x~aust Fp, n~e~ Gasoline Exhaust Pestled e s,,H edok:~,~es Tex~le X-rays~Radio~ M~teda~ REMARKS: MISCELLANEOUS: |o Where were you born?, ~ ~ were youtpaten~ born? b) What warn you~ date,J o~ ~en,,k:e? ,to c) V~m did you serve? 9. What is lt~e most upaettlng event tl~t happened to you in about the last fiv~ yean? [] None 0 0 ,-,.I 0 O~ I I ! I I I I 1 I / I I I

I I I I I I I I I I I I I I I AMERICAN CANCER SOCIETY ~.~=.,,~. ~~ CANCER PREVENTION STUDY I! °'*. ~. I Reseatctmr No. QUESTIONNAIRE FOR WOMEN Unit No. Family No. 0a~e: 1. Name: .... 2. Date of birth: Month Year,~ 3. How o,~d am you now? , 4. Current weight with indoor doming: .... 5. We{ght 1 yesr a~jo: ..... 6, Height (without shoes): .... ~, in. FAMILY HISTORY (IN RELATION TO CANCER): I. RII in ~ follov~ table as corrcletely as poss~e for pamnt~, brothers and sister¢ IF DEAD. OlD THIS PERSON EVER ! tF'YES.- HAVE CANCER? SPECIFY (Ci~e O~e) TYPE OF CANCF.R Yea NO Yes No Yes No Ve= No Ye= .No ., Ye~ No b) How ofd was yo~ falt'~'? Whst pad of your body? 6. How many times have you had co~s o~ flu in the past twelve months?, 0 0 0

CURRENT PHYSICAL CONDITION: t. How much exercise do ~0u get (work or l:Way)? [] None [] Slight [] Moderate r] Heavy 2. On the ~rage. how many hours do ~ sleep each night? 3. On the average, how many times a month do You have insomnia? . ~ , r'l None 4. Within the last twelve montt~, ha~,e you noticed: a) A lump of thickening in your breast? [:]Yes nNo b) An unusual disc~e Imm your breast? [] Yes O No 5. Oo you notre pains in your legs whe(1 you walk whk:h go away w~nen you rest? []Yes [] No ff "yes," how many year~ have you had these pain~? 6. Am you sickatthe presenttJme? []Yes ONo If "yes," with what disease o~ conditJoo? MENSTRUAL AND REPRODUCTIVE HISTORY: 2. W'nat is your current menc~3~saJ status? 8S~g m~ men.~-ua~ng In ~se ~ ~t m~au~ b) ~ ~ (~) ~ ~u~ humor of d~s ~ 4. ff past menopause: a) Was your menopause: [] Nmur=l [] Artff'w=i~d b) Age when periods stopped completely? .... C) Did you hav~ excessive bleeding dudng menopause? [] Yes [] No 5. Hm~e you ever had or t~ed to ha~e ch~ren? nY~ •No If "no," skip to question g. 6. Have ~ e~r h~d difficulty beccx~ing pregnant? []Yes ONo if'yes," what was the mason? 7. How many times have you been pregna~? a) ~ age at ~0ur ~ pm<Jnancy?. b) Your age at ~:~Jr first live birth? ¢) Number of children born aJiv~?. d) Numbs' of stillbirths (carded 5 months or mo~)?. e) Number of misca.qiage= (canted less than 5 monthsl?. 8. Were you ever given DES (Dieth~lbestmO to peevent mLscantage? [::] Ye~ [] No =) At what age did you take it? ..... b) For I'mw malty months did you take it? . 9. Birt~ control meth<x:ls: Intimate your age when first used and number of years o! use. NONE OF THE ABOVE ~ Have YO~ ever taken ora~ ¢on~ptives (birth cont~ p~P=)? ~ ~ ~ No If~:~ ~q~ 11. ~e~ fi~t~m? b) ~ ~y ~ d~ ~ ~e mem?.~ 0 0", 0 0 C I I I

I I I I I I I I I I I I I I 3. Current and ex.smoke~: a) Do (did) you inhale? [] No, never O Slightly [] Moderately I-i Deeply b) Fill in the following information foe:. 1)The first beand smol~ed regularly. ~ 2}The brand of cigatelte smoked fo~ the longest period of t~me. DIET: 1. Onme average, how many days per week do you eat the following foods? (If less than once =week, Beef Raw vegetable= Pod¢ Canto Uve¢ Citrus fruits/Juices Ham Spaghetti/M~ca~ Fmh Wn~e r~e Smoked meats Wb~e bread~ls/ Frankfu~ecs/ Biscu~ Sausage Brown r~ce~Wh~e BuYer whea~Ba~ey Cheese Po|atoes Eg~s Oa~rnea~S~%-dded Green [ea~ wheat4~ran Tomatoes Co~d (Dry) cereaJs Brussels spmuts~ Chocolate 2. How many day~ a week do you eat the following fried foods? Fried eggs ~ Fded hamburgers Fded I~co~~ ~ beef Fried chicken~'~ Othe~ fried foods Fm~:J~ hies DO NOT EAT FRIED FOODS [] 3. Do y~u eat = vegetarian diet? l"l Yes O No If '~jes: what P/pc and foe how many yean~? 4. Has there been a ma/~or c~an~e in your diet in ttm last I0 yea~? [:::)Yes C:}No 5. a) DO you now o¢ h~ve you ever added attiflciaJ sweeteners (saccharin or cyclamates) to coffee. tee, o¢ O~ler drinks o¢ food? OYe~,c~m~W O~ ONe~,r amount p~ ~ and fo~ how long. Packets: No. per dW.~-..---.. Year~ Oro~=: No, per day,~ Yearn Tablets: No. per day Ye a.,~ , 6. DO ~:~u get your drinking w-~ter from: 0 City ~l:~ly [] Pnvate well [] Other (spec~) 7. ~ ~ add ~y su~s lo ~ ~r ddn~ 0 Y~ 0 NO 8. H~ m~y ~, gl~es, ~ ~ ~ ~ ~- ~es ~ ~ u~ly ddnk a day, ~ f~ ~ ~y ~es ~t~ ~[emi~(~s~m milk) ~et ~ ~ ~ ~ tee N~tet ~ ~ ~et ~ ddnk ~r H~ MEDICATIONS AND VITAMINS: 1. How many times in the last m<mt~ have ~tou used the following ~ how long hm~ you u,s~d them? If none, write O; If used orgy occadona~wdte 1/2.) Aspirin, Buffedn. Anacin V~tamin A ~rda,min C V'rtamin E Multi-Vffamins Brood Pr~.sure pills Diuretics (water pills) Tnymid mea~.~gons Heart medications Valium Libdum Proscription sleepin~ pdls Tagamet (fo~ ulcers) Other.

OCCUPATIONS: !. Wh~t is your current occupation a~d what am your ~uttes? ...... How many years: 2. I! re|i~ed, what w'~s your last occupation? 3. What o~etjob have you heal fo~t~a k~gest period of tLme?, paid ~ . ~oluntee" wo~ .... housewod~. 6. In you~wotkor d~ly life.am (were) ~ ~ulady REMARKS: MISCELLANEOUS: 1. Wt~ere were you born? 2. W1~em wet'e your pa~nLs boon? Fathe~ Mother: I I I I 1

I I I I I I I I I I I I I I AMERICAN CANCER SOCIETY CANCER PREVENTION STUDY' II INSTRUCCIONE$ PARA LOS ENTREVISTADORES INSTRUCCIONES GENERALES: ln.scdba a almdedm de diez fatnilias: Pot famillas se er~ande hogares deride hay rods de urm persona viviendo juntos como una familia. Cada familM debe re'me per Io menos una persona que teng~ rods de 45 afros. Pot favor haga un esfuerzo pot encocfaaf familias con pemonas entxe las edades de 50 a 50 ahos. Inscfiba sdamente aquellas familias las ¢ua~es usted esM bastan, te segum clue van permanecer en el mismo vedndado durante los p~ximos sels ~ $i usted puede inscribir rnds de" diez familias, pot favor h~jaJo. Pata ayudar a exlN'¢ar el propdsito y el plan de este estudio deje el pantleto "Cancer Pm.~=nlk~ Study II---Hoja Infon'na~va" con cada familia que ustedinscdba. En cade una de las familias que usted inscdba, pida ClUe cada miembro que sea mayorde 30 Ilene o co~teste et cuestionado, los adoque en un "Solxe Cocd~denciel," Io cierre y se Iode,,uelva a usted. Debido a Io extenso de este estudio es necesado identiftcaf carla ~ c(m una sede de nt~metos. Esto se expiica en el I:~UTab n~mem 3, abajo indicado. Pot re#of siga las ins~'tx:docms cuidadosemente. Despuds de recoger los cuestJonarios, Ilena el folleto de cuabo i:~ginas °Usta de Farnirm y Inscdtas." Induya en ~te el hombre y la direccidn de una persona que o:mozca a la ma~a de familias lnscdtas y que pueda reemplazado durante los p~xJmos seis atk~ si fuera ne:esario. Cuando dsto estd terminado meta los "SoOms Con~enda~eS" (con los cueslJonados campletos) en folleto "Usta de FamOus y Personas k~scdtas," asegt~m~o cort una gorna el/L~ca y co~elo sobm grande. Entregue todo el mateda~ terminado seg~n las ins~JCdOrmS que se INSTRUCCIONES DETALLADAS: 1. Revise et paquete pata aseguratse de que cordJenelo dguiente: a) sufidentes cues~o~ados paca hombres ('~npt~o eft ¢olo¢ azu0; b) sufldentes cu~ados para mujems ('m, tpmso an color INanco); c) suficient~s "Sdo~ Co~fKk, nda~es;" d) un folleto de cuatro I:~ginas "Usta de Familias y Personas Ina:dtas;" y e) suf~derttes'Hojas 2. En la Oltima pdgir~ de este folleto de instnx:cio¢~ enumere las fame'ms (hogares) en las que usted que riven iuntas como una familia y tambid, n inc~ a persocms soiteras que riven =dmt Vis~te a cada farn~T=a en su lista e inscdba so(amente aquellas que usl~d piensa estar~ an el dma durante los prbximos seis a/tos. Pr=da ClUe cada miembm mayor de lminta a/~o= liege el ~. No exduya auna familia si uno o dos ndembms se niegan a Ilenaf o no lenan el cu~ despuds que otms rdemb~:~s de la fa.mlia Io hayan Ilenado. 3. Para fac=Tt~r la ident~:addn a usted se le ha asignado un Ndrnero de Divisidn, un N(mero de Unidad, un Nt~mem de Gn~po y un N0mero de Enlz~wistado¢. Cop{e todos esos n0meros en todos ndmeros de ident/licaddn, su nontze y direcci~, yel notable y dimodd~ de un subs~tu~o que conozca a la rna, F~'a de las familias clue usted ha inscdto, en la parte de a~dba del f~eto "L~"la de Familias y Personas Inscdtas."

As~gnele un N0mero de Familia, a c~da f~milia que inscnba, siendo segunda Familia No. 2. etc. Aden',ds as~gnele un Nt]mem de Persona, a c~da persona que insc~ba en cada farnilia~ siendo un miembro de la t'amilia Pe~--=ona No. 1, otto siendo Perso~ No. 2, etc. Pot eiemplo, la pdmer~ familia (Familia No. 1) puede componerse de! Sef~r y Sef~c)ra Ldpez, su hiio de 35 a~os J(xge Ldpez, la suegra del Ser~ Lc~z, ia Sef~ora Rivera y un amigo, el Ser~" Ricardo Martfnez. Entonces, papa la Familia No. I, el Se~r Ldpez es Persona No. 1, La Se~ora Lbpez es Persona No. 2, Jorge I.bpez es Persona No. 3, La Se~,ora Rivera es Persona No. 4. y el Se~or Mattfnez es Persona No. 5. Oespuds, el Se~or y Set~'a Brown pueden set la Famitb. No. 2; dendo el Set, or Brown Persona No. 1 y ia Se~ora Brown Persona No. 2 en la familia. 4. Cua.ndo una persona acepte Ilenar et cueslionado, escdba el hombre de dl o ella y todos los r~meros de identificaci~n (indu)~=ndo elNdmero de FamEa y el N0mero de Persona) en la paste de arriba del cuesttonado. Tambidn escriba et nombre de ~l o ella y la dlreccl6n y todos los n~nems de iden~cack~ en e~ "S4:~:~e Cont'K~endal." Entrdguele el ¢uestiona,'b y d "Sobm ~ncial" ai patr~ante. El cuest~ortatto estzi, diserlado papa set Ilenado pot la persona y las contestaciones son ¢onfldenclales. Pida que el pa~pante Ilene et cuesdonado y luego lo meta en m sobre y lo selle. Usted es responsabte de mcoge~ ~s sobres sellados. Usted puede esper~ rrientras el pa,"tidpante complete e{ cues~ona~'io o, d usled Io ~'effere, puede dejar el cuestionado y ~resar rnds tarde a mcogerlo. 5. Trate de inscdbir a tc~as sus famil'~s y de recoger los cuesSonados co~etados en un pedodo de 6. Despuds de clue usted haya recogido los cuestlonados detodos las personas que usted ha inscrito, ya terminados, Ilene la "Lista de Familias y Personas I~," seg0n las instrucciones dadas en este folleto azuL 7. De.spuds de que haya completado todo, meta'Los Sobres ConF~enci~es" junto cc~ el f~lleto "Lista de Familias y Personas Inscdtas," asegum todo con una goma eldst~ca y pc~jaJo en el grande y devudlvalos seg0n las instn.cciones recib~as. OBJECTIVO Y PLAN DEL ESTUDIO: E~ primer Estudio Sobre la Pmvencidn de la Sociedad Americana Contra et C.dmce¢ se Ile'~ a (:~bo durante un per~odo de 13 afros, desde 1959-1972, y nos ayud~ a ide~tif~ca~ un nOmem de fac~res relacionados con e~ desano~lo de! cdncer, l~e hecho, mucho de Io que conocemos hoy sobre causas del cdncer ha. eurgido de @stos estudos epidemiok~icos. E] ~o N~'nem 1 de 1,1 ~n del Cdncer, pot ejemp~o, establed~ clue el ~umar dga~dlbs es una de las pdncipa~es ¢ausas del cdncer del pulrndn e impC~ a~ uso de~ taba¢o en el desarrollo de otros tipos de c:dncer y en las enfermedades del cocaz~ y vias respi'a~odas. Otms es~Jdios epidemic~icos hart vinculado aJ cdncej" de la piel a demasiada e~ • los F~oX, arsds-~:o o dertos tJpos de bmas ¢dtcer de la vejiga, a lTabajadoms expuestos a dertos pmcludos qufmicos y expos~ durante largo l~ernpo alas 5b~as de a.sbestcs. Esta~ so~ aJ~unos de ~ ~ ambiences clue pueden c~usar ~ Es ~amente a tra~ de la observ-~c~n de un ~'nplio r~mero de persanas durante un largo pe~od~ de t~em~o, coco pS~neamos hair en e~ Estudio N~mero II Para la Pmvenci~n del ~, que podemo~ descubdr muchos ob~s lactates y determine" cudles son perjud~:dales paJ~ la s~lud y cu~d~s no. En el Estudio II Para la l~nci~'t de~ ~__,dnce~ vamos a enfocar nue~tra atenc~ l~a los c~os que hs~ ocu .n~do desde nuestm pdme~ estu~o en nuestm esdo de rids, los pmduc~os que us~mos y en el arnbiente de nuest~o hogar y l~gar de em~o. Re~enter~ente, h~ habk~ un grin iriter~s en determinar el efecto de la s~.,arina, l~ntes I~= el cabello, ~ptivo~ drogas y medicamentos. El efecto de la expodc~ durante I I I I !

I I I I I I I I I I I I I contaminad0n del aim y del zgua, y los carcinbgenos en los lugams de empleo tarnbidn necesitan set cuestionados. Ei p~iblico y lacomun/dad cient~ca desea encontra~ la tazbn p<~-a el aumento en los ca.sos de cdnce~ en la poblaci~n negra y se~'la~ Io~ espeda~es de.sgos de cdncer ent,~e ot~.s minodas. El plan del nue~o Estudb Sobre es inscdb/r a rods de I l/empo. Como entrevistador toluntado, usted serd de gran ayuda en recopi~ infon'n,=~Z~n inves~a}Jva vita~. Manteni~=<~ose en contmto con las pemoc~s que ha inscdto e info~Tnando sobre ellos cada dos af~s, habrd suministrado a nuestros estadfsl~cos de sa/ud con informac~n sob~ cbmo los est~os de vida afectan la salud y qu~ factores aumentan o disminuyen las opodunidades de adquidr cdncer y (:~,ras enfermed~es. Este t~po de estudio ~ument~ nuestm co~dmiento sdore el cdncer y nos pecm~¢d, idenE~r aquellos factoms que c~usa~ cdncer y que pueden set" corttm~doso as( como aque(k:~s que no y sahrat miles de v~das. ALGUNAS PREGUNTAS QUE LE PODRAN HACER LAS PERSONAS QUE USTED INSCRIBA: R /.Po~'qud fur escogido par~ este estud~o? R. Neces~J'nos inscdbir un= ampii~ muesCa de diferente p~b~oo: personas de d~ferentes ed~des, ¢'eas geogrd~cas, razas, relig{ones, hdbitos, expo.s~k~tes y est~los de encontrammos cu~des grupos tienen de.~os mCs aJtos de contr~et" cdncer y cud/es los mc'Ls halos. R z, Est~ interesado ma~:~rrm, nte en personas con FL No, estamos interesados en todas las perso,'~, aquellas que est~ en buer, a saJud, asf como aquellas clue t~enen o hart tenido ¢dncer. R Mi hip de 25 afros ~ ccr~migo. ~,Por qu~ usted no de.sea que R. Estamos exduyendo m pe~ortas menoresde 30 afros pocque ellos no h~n ddo expuestos mbs frecuenda de( ¢dnce~" gene~mente aume"~ coot Im adad y no habd~ sufcie~te informaci&'~ para • estudia~ s~ inscdbimos I:~'~onas menores de 30 afros. R Nosotms sabemos ya que el fumar dgamlos causa ~ ~.Por qud necedtamos otto estudio? R. Los dga,'Tillos fumados ~om p<x rods de dncuenta millo~es de personas son considerabiemente diferentes de los fumados en la dpoca de nuestro pdmet esC, dio. Necesitamos determir, cr dga~dllos bajos en brea y nicot~na hart afect~o substandaknente los riesgos de saJud. Tambidn estamos investigando los efectos del fumar cigamllos en e~ ambiente de lugams de empleo y los posibles efectos de saJud del fumador de "segunda-rnano," ~'to es, el humo inhalado pot pe~o~'tas que no fiJman. R /.PO~lU~ me preguntO p(xmi n0mero de Seguro SociaJ? ~,No es eso ilega~? FL Damos su n0mero de Seguro S<x:ia~ es ex~ctamente vo~untado. AJ hacodo, nos ab, o~ra~ usted mucho l~empo, esfuerzo ydinem aJ veri~:a" nuestros a~chi,ms rods tarde (especiaJmente para persons con bs mismos hombres). Casu~mente, no es ilegel pedir su nOrnem, es ileg~l exlgfr,~do. R ~,Se mantendrd confidential I~ inf~b't en el cuestionado? R. S(. Ser~ ut~liz~dm solamente papa los pmpbsitos de la inves~o Nunc~ daremos informacidn sobre ningun~ persona en p~rl~cular y no dammos direcciones a ninguna mgencia po~ ningtin Ixopd~sito, cuaJqu{em que ~ste sea. ~.~ 0 CO 0"~ 0

2063620716

Appendix E. Bibliography of Epidemioiogic Studies of Lung Cancer and ETS L.is~..~ by clm:s'.olo~ic !. Hh-ayama T: Non-~noking wive~ of he•vy unoke~ have • higher incidence of lung caace~. study from Japan. BMJ 1981; 282: 183-185. " ' 2. Trichopoulos D. Ir~landidi A, Sparro~ I~ MacMahon B: Lung cancer and passive J Cancer 1951; 27: 1-4. 3. Garfinkel I.z Th'~e ~endx in lung cancer mortality among non-m~oken and a no~ on passive ~moking. J]~C~ 1981; 66 (6): 1061-1066. 4. Chang WC and Fung SC: Lung cancer in nonosmoke.~ in Hong Kong. I~ Grundmaan (~): Cancer Campaign Vol. 6. Cancer Epidemiology. Gustav ~her Vedag. Sm~zga~ H York. 1982, pp. 199-202. 5.Trichopoulos D. Kalandidi A and Spa-ms L: Lung cancer and passive smoking:. conclusion of Greek su~iy. Lancr~ 1983; ii: 677-678. 6. Con-ca P, Pickle LW0 Fomham ET, IAn Y, and Hacnszcl W: Plosive smoking and lung cancer. Lancet 1983; ii: 595-597. 7. Kabul GC tn¢[ Wynder EL: Lung car, oct in no~smoke~. Cancer I298~; 53: 121a,-I22L 8. Hkzyarna T: Cancer mortality in ~nsmoking wemcn w~h unoking husbands on large-scale cohor~ seedy in lapaa. P~v Med 1984; 13: 680-690. 9. Buffier PA. Pickle LW. Maso¢~ TJ, C.~ntant C.: The causes of lung cancer in Texas. Mkzell M and Cot•ca P:. Lung canc~:, causes ami ~tion. Vedag 198,t, pp. 83-99. l(}. Sandier DP: Passive in adukhoo:[ and, ca•leer risk. AlE I985; 12I (l): 37-48. l 1. Garfin~l L, Auerhach O, and Ioul~rt L: Involuntary smoking tnc[ lung ~ conm31 study./NCI t985; 12. Wu AH, Hen~c~on BE, Pike MC, Yu MC: Smoking a~ui othe.r ~ist" favors for lung cancer in women. £bfCl 1985; 74: 747-75L 13. Akiba $, [Cuo H, Blot WI: Passive smoking ~ lung canc~ among $apanese women. Can Rex I986; 46: 14. Dalagcr NA. Pickle LW, Mason "rl, Cot•ca p~ Fomham E'r, $temhagen A, Buftler PA, Zicglcr RC., Fraumuni JF Jr.: The relation of passive smoking and lung ca~ce~. 15. L~ PN. Charal~-rltin l, and And~'~ MR: Relationship of pas~ve smoking to ~ of. hmg cancer and od:er ~rnoklng-associatcd diteas¢~ Br.[ Cancer I986; ~4: 97- O I

I 16. Gao YT, Blot W~'. Zheng W. ershow AG, I-Isu CW. Levin LI, Zhsag R, Fraur~ni/F Jr.: Lung Cancer among chinese won~a. UC 1987; 40: 604.609. 17. Brownson RC. Reif IS. Keefe TI, Fergwon SW. and Priml IA: Risk f~-'~o¢~ adeaocarcinon~ of the lung. AJE 1987: 12~ (1): 2~-3~. 18. Koo LC. Ho/I-IC. Ho C: Me-cun:mcn~¢ of passive s~noking and ~dmate~ of lung cancer risk among non-smoking chine.~ ferrules./5C 1987; 39: 162-169. 19. Per~hagcn G. Hn~bec Z. ~nd Svens~o~ ~: l~ivc ~ra~king and lun~ ~ Swedish women. ~ 1987; 125 (I): 17-2~. 20. Hutoble CG. Samet ~v[. P~tha~¢ DR: ManY,go to s ~noke~ and lung c~ncer r~k. 1987: 77: 595.602. 21. L~m TH, Kung [TM. Wong CM. I.am WI~ Klccvcas ~ Saw D. H~t C., Sacvirame I'L Lain SY. Lo KK. and Chaa WC~ Smoking, p~.~-ive smoking histological type~ ~n lung c~cer in Hoag Kong Ch|ne~ women. Br~ Cancer 1987; 56: 6"/3-678. 22. ~ TH, Che~g ~ Pa~iv~ ~-aokiag i~ • t'Lck fac~orfer lung cancer in n~ver ~noking wom~n in Hong Kong. Smoking ~ad He.~lth. F.,L~vie, r. 1987, pp. 279-2gL 23. C~ng GY, Ling ZH. Wu GL: On ~he r~la6onship bctw~n sraoldng and female lung cancer. Smoking and Health. El~vier. 1987, pp. 483-486. 24. Inou~ R, Hirayama T: P~sive ~raokin~ ~d lung c~ace, r in women. Smoking Health. Elsevier. 1988, pp. 283-285. 2.5. Kamda H, Mikami R. Konishi R~ Ko~ma Y, a~i N~rita H: Effect o~ pa~iv~ in lung cancer de~elopmem in worn~n in ~he Hm-a ~gio¢~ Cran Ho P, insho 1988; 21-27. Tr'~a~|ated from texx in ~apan~ by M~. Yumt Imai. 26. Shimizu It, Modshita M, Mizuno K, Ma~uda T, Ogura Y, Santo M, Ni~himura M, Kunishima K. Yamamo~ M, His~raichi $, Tominag~ $: A ~-con~rol ~'udy of lung cancer in non,rooking women. Tohok'~ I exp Meal 1988; 134: 389-397. 27. Butler TL: The r~ladon~h~p of pasxiv~ smokin~ ~o v~riou~ health ou~:om~ Seven-Day Adven~u in Califotoim Dissera~ion. L~ An[~l~: UCLA. 1988. 28. Svea~on C, Pe~hag~a G, KIotoinek h Smoking ~d pa~iv¢ xtooking in lung cancer in wocnen. Acu~ Oncol 19g~ 2g: 62:~4~29. 29. Hole DI, GilIis CR. Chopra C, I-~wthom~ V'M: Passive smoking and ca~ior~spira~oq, health in a general populafio¢~ in ~he w¢~ of Scodaad. Br Med ~': 1989;, 299: 423-427. 30. Wu-WiHiam~ AH, Dai XD. Blot W, Xu Zy, Sun XW, Xiao HP, $~on¢ F~a[ YP, Er~how AG. Sun ,~, Fraumoni 1/: It., and Head~n'~on BE: Lung cancer among women in unrth-ea~t China. Brl Cancer 1990;, ~2: 982-987. 0 0 I 4 I ! I I I I i I

I I I I I I I I I I I I I I 3 L Janerich DT, Thompson WD, Varela L, G~nw;~d P. C'homst S, Tuc~ C, Z.~-nan MB, Melamed MR, Kiely ~ McKneally MI:: Lung cancer and exposure ~o ~obacco in the household. NEIM 1990;, 323: 632-636. 32. Sobuc T: Assoc/afion of indot" air pollu~on and lifestylc w~th lung cancer in Japan. LIE 1990; 19 (3) Suppl. : $62-$66. 33. Ka/andidi A, Ka~ouyunni K, Voropoulu N, Basta~ G, Saracci R, Trichopoulos D: Passive smoking and diet in the etiology of lung cunc~ among no~-smokets. Cancer causesanrlcontrol 1990; I: 15-2L 34. Liu Z, He X, Chapman RS: Smoking and other risk factot~ for lung cancer in Xuunwci, China. D'E 1991; 20 (I): 26-31. 35. F-ontham ET, Corr~a P, Wu-WiIILIn'd A, Reynolds P, Gteenberg P~, Bufflcr PA, Chen VW, Boyd P, Ale-man T, Austin DF, Liff JM, G~eenberg DS: Lung cancer in nonsrnokiag women: t multicenter case.conm31 study. Cancer EpidcrniololD', Biormrk~ and Prevent/on 1991: I: 35-43. 34. Kabat GC: Epidcrrfioingic studies of the relationship between passive smoking and lung cancer. Pn:scnu:d at the 1990 annual yarner ~po~ meet/ag of the Toxicology Form= in Washington DC. 35. Rcif.IS, Oglivc GK and Harris CK: Passive smoking and can/no lung cancer. AJE 1992; 125 0): 234-239. 36. Trichopoulos D, Mollo F, Tomatis L, Agapkos E, Delscdirnc L, Zav~tsaaois Kala~didi A, Kat.~ouyunni K. Riboll F., ~d Sancci R: Active and paxsive smoking and pathological indicators of lung cancer risk in an autopsy study..IAMA 1992; 268: 1677- 1701. 37. Stockwcll HG, Goldman AL, Lyman GH, Noss CI, Armstrong AW, Pinkham PA, Candslora EC, and Brusa MR: Eav/rmmen~ tobacco smoke and lung cancer risk in nonsrnoking women. JNCI 1992; 84: 1417-1422. 38. Brownson RC, Alavanja MCR, Hock IST. Loy TS: Passive smoking ~md lung canc~ in nonsmoking women. AJPH 1992; 82: 1525-1530. 39J..iu Q, Sasco A,I, RiboH F., and Hu MX: Indoor air pollution and lung cancer in Guangzhou, People's Republic of China. AJE 1993; 137: 145-154. 0 0

2063620720