Tobacco Documents Online

PRIMARY EPIDEMIOLOGIC STUDIES ON SPOIISAL SMORING AND LUNG CANCER Introduction Currently, 34 epidemiologic studies examining lung, cancer incidence and spousal smoking have been published or are`available as dissertations or conference presentations.1-34 The material in Section A of this notebook is discussed according to geographical area: United~ States (Tables 1 and 2), Asia (Tables 3 and 4) and Europe (Tables 5 and 6). For purposes of comparison, the relative risks (point estimates) given in the tables are the overall point estimates for spousal smoking reported in the papers. In some cases, the risk in~ the table was selected from numerous point estimates presented in the paper, based on different definitions of exposure, break-down of the sample by histological type, etc. Tables 7 and 8 summarize the workplace and childhood exposure data currently available. Brief synopses and copies of the papers associated with these studies follow this introduction, at Tabs 1 to 34. For several of the more recent papers, copies of letters to the editor concerning the study may be found following the primary paper. The extensive criticisms of the Hirayama and Trichopoulos, et al., papers, however, are discussed in Section C of this notebook. The copies are highlighted in yellow for useful information and in~ blue for negative statements.

United States Studies Fourteen of the available studies on spousal smoking and lung cancer in nonsmokers (two cohort, twelve case-control) were conducted in the United States (Table 1),3,5,7-9,11,14,16,24,25,30- 33 None of the overall relative risks (RR) for spousal smoking reported in these fourteen studies is statistically significant. The United States Environmental Protection Agency's 1993 Risk Assessment on ETS relied heavily upon eleven of the U.S. studies in reaching its conclusion.3,5,7-9,11,14,16,24,30,33 Although none of the studies originally reported a statistically significant overall risk estimate for spousal smoking, EPA used these studies to arrive at its conclusion that ETS exposure was associated with a statistically significant risk of lung cancer in the United States. In its analyses, EPA recalculated 90% confidence intervals for the risk estimates, instead of adopting the more commonly used 95% confidence interval. At 90%, one study, by Fontham, et al., had an overall risk estimate that was statistically significant. Nevertheless, ten of the eleven studies cited by EPA were compatible with the null hypothesis of no association between N spousal smoking and lung cancer risk. Q ~ W Cll Among the more recently published papers, the paper by ~ Janerich, et al., is based upon an unpublished dissertation by ~ CT~ - 2 -

Luis Varela.24 The Janerich, et al., paper discusses a subset of Varela's case-control study, and reports no statistically significant increased risk for spousal smoking, workplace exposure, or exposure in social settings. (it does, however, report a statistically significant increased risk for exposure during childhood (see below).) Overall, the Janerich/Varela study is important because of its large size and appropriate study design. Three major new case-control studies, conducted in the United States, were published in 1991 and 1992. 31-33 Fontham, et al., presented a preliminary report on an ongoing multicenter case- control'study.31 While the study design~includes measures designed to minimize methodological concerns, the study is nevertheless incomplete: all data have neither been gathered nor fully analyzed. The Fontham, et al., study reports statistically significantly elevated risk estimates for adenocarcinoma, but not for the overall index of spousal smoking. A follow-up article claims that a number of confounders have been considered and dismissed by the authors. A relatively small case-control study by Stockwell, et al., appeared~in 1992.32 In contrast to Fontham, et al., this study reported generally lower risk estimates for adenocarcinoma than for other cell types, a striking example of the inconsistencies among the reported results of these studies. Many risk estimates 3

in~the Stockwell, et al., study, including the overall risk estimate for spousal smoking, were not statistically significant. The third recent study, by Brownson, et al., is a very larg,e case-control study, conducted in Missouri.33 The authors of this study report no statistically significant risk estimates for any "'quantitative'"' estimates of ETS exposure. Unlike the Fontham, et al., and Stockwell, et al., data, Brownson, et al., reported no statistically significant risk estimates when their d!ata were analyzed by cell type.

TABLE 1 -- UNITED STATES STUDIES OF SPOUSAL SMOKING AND LUNG CANCER IN NONSMOKING WOMEN STUDY NO. OF CASES1 OR2 95$ CI Brownson, et al., 1987 10 1.68 (0.39-2.97) Brownson, et al., 1992 218 1.0 (0.8-1.2) Buffler, et al., 1984 33 0.78 (0.34-1.81). Butler, 1988 4 2.04 (0.54-7'.65). Correa, et al., 1983 14 2.07 (NS)3 Fontham, et al., 1991 264 1.21 (0.96-1.54). Garfinkel, 1981 88 1.27 (0.85-1.89). Garfinkel, et al., 1985 134 1.22' (0.97-1.71) Humble, et al., 19874 16 1.8 (0.6-5.4)5 Janerich, et al., 19904' 129~ 0.93 ( 0. 55-1. 57'). Kabat,, 1990 35 0.90 (0.46-1.76)~ Kabat & Wynder, 1984 13 -- (NS ) 3 Stockwell, et al., 1992 2106 1.6 (0.8-3.0) Wu, et al., 1985 29 1.2 (0.5-3.3) 1. Number ot nonsmoking lung cancer cases married to smokers, and used in spousal smoking analysis. 2. Odds ratio for overall index of spousal smoking, as reported~ in original publication. 3. OR and/or CI not given; reportedly not statistically significant (NS). 4. Data are for males and females combined. 5. 90% CI, as reported in original publication. 6. Total number of cases; numbers not given~for individual analyses. a ~ CA ~ N Ul N cc

TABLE 2 -- COMMENTS ON UNITED STATES STUDIES OF SPOUSAL SMOKING AND LUNG CANCER IN NONSMOKING WOMEN STUDY COMMENT Brownson, et al., 1987 Brownson, et al., 1992 Buffler, et al., 1984 Butler, 1988 Correa, et al., 1983 Colorado; adenocarcinoma only Missouri; large case-control study; partial NCI funding,; not included in EPA Risk Assessment Texas; case-control study; no statistically significant risks reported for indices of ETS exposure California; Ph.D. dissertation; never published; small sample size; deals with specific religious group, the Seventh- Day Adventists Louisiana; large case-control Fontham, et al., 1992 study, but extremely small sample size for ETS analyses Five cities;; report of on-going study; large sample commendable design size; when completed; high proportion of adenocarcinoma Garfinkel, 1981 Part of American Cancer Society prospective study; large study, but few deaths cohort among nonsmoking women; data contrast with Hirayama's data from~Japan Garfinkel, et al., 1985 New Jersey and Ohio; numerous risk estimates presented; strong indication of respondent bias between spouse and children Humble, et al., 1987 New Mexico; small sample size

TABLE 2 -- CONTINUED Janerich, et al., 1990 New York; large case-control study, based on data from 1987 Varela dissertation (itself unpublished); many risk estimates, only that for high exposure during childhood statistically significant Kabat, 1990 Report of on-going American Health Foundation study, presented at scientific meeting; no statistically significant risk estimates reported;; study appears to be well-d'esigned Kabat and Wynder, 1984 No significant differences between cases and controls regarding ETS exposure at home; concludes with detailed discussion of epidemiology of ETS Stockwell, et al., 1992 Florida; small case-control study;; results contrast with those of Fontham, et al.; all risk estimates were not provided in publication Wu, et al., 1985 California; adenocarcinoma only (too few small cell cases to analyze)

Asian Studies Approximately 14 epidemiologic studies on spousal smoking and lung cancer in nonsmokers (one cohort, 13 case-control) have been conducted in China, Hong Kong, and Japan (hereafter, "Asian studies") (Table 2),1,4,10,12,15,17,18,2'0-22,27,28,29,34 Of this group, several studies report statistically significant overall risk estimates. However, none of the reported relative risks is greater than 2.5; relative risks under 3.0 have been described as "weak" (see Criticisms section~in this notebook). Of particular interest is the 1990 paper by Wu-Williams, et al., conducted in northeastern China.28 This large case-control study reports a statistically significant negative risk associated with ETS exposure. Other factors (particularly indoor air quality) were reported to be associated with an elevated risk of lung cancer imthe Wu-Williams, et al., study; such confounders were not always accounted for in the other Asian studies (see section on Confounders in this notebook). N © N ~ ~ ~ Cli 1~J

TABLE 3 -- ASIAN STUDIES OF SPOUSAL SMOKING AND:LUNG CANCER IN NONSMOKING WOMEN STUDY NO. OF CASES1 OR2 95$ CI Akiba, et al., 1986 73 1.5 (1.0-2.5)3 Chan & Fung, 1982 34 -- (NS) 4 Du, et al., 1993 47 1.19 (0.66-2.16) Gao, et al., 1987 189 0.9 (0.6-1.4). Geng, et al., 1988 34 2.16 (1.03-4.53)(SS)5 Hirayama, 1984 163 1.45 (1.04-2.02)3('SS) Inoue & Hirayama, 1988 28 2.25 (0.91-7.10) Koo, et al., 1987 51 1.64 (0.87-3.09) Lam, et al., 1987 115 1.65 ('1.16-2.35) (SS) Lam, 1985 1636 -- -- (SS) Liu, et al., 1991 45 0.74 (0.30-1.96) Shimizu, et al., 1988 90 1.1 (NS) Sobue,, et al., 1990 64 0.94 (0.62-1.40) Wu-Williams, et al., 199& 205 0.7 (0.6-0.9) 1. Number of nonsmoking!lung cancer cases married to smokers, and' used in spousal smoking analysis. 2. Odds ratio for overall index of spousal smoking, as reported in original publicatiom. 3. 90% CI, as reported in original publication~.. 4. OR and/or CI not given. NS= Reportedly not statistically significant. 5. SS = Statistically significant. 6. Total number of cases; numbers not given for individual analyses.

TABLE 4 -- COMMENTS ON ASIAN STUDIES OF SPOUSAL SMOKING AND LUNG CANCER IN NONSMOKING WOMEN STUDY COMMENT Akiba, et al., 1986 Hiroshima and Nagasaki, Japan; study of atom bomb survivors Chian & Fung, 1982 Hong Kong; small study Dui, et al. , 1993 Guangzhou, China; small preliminary report study; Gao, et al., 1987 Shanghai, China; looked at a number of potential confounders Geng, et al., 1988 Hirayama, 1984 Inoue & Hirayama, 1988 Koo, et al., 1987 Lam, et al., 1987 Lam, 1985 Liu, et al.,, 1991 Tianjin, China; small study;; limited information available Japan (six prefectures); large cohort study, first published in 1981; heavily criticized for improper age standardization and other flaws Kamakura and Miura, Japan; small study; limited information available Hong Kong; many papers published on these data, some including interesting data on potential lifestyle and dietary confounders Hong Kong; problematic method of control selection Hong Kong,; unpublished dissertation, only some pages made available by University; adenocarcinoma only; see Lam & Cheng paper for some details Xuanwei, China; small study; presence of at least one smoker in household used as surrogate