Tobacco Documents Online

COMMEN'rS OF R. J. REYNOLDS TOBACCO COMPANY ON HEALTH EFFECTS OF PASSIVE SMOKING - ASSESSMENT OF LUNG CANCER IN ADULTS AND RESPIRATORY DISORDERS IN CHJLDREN (EPA/60015q~:)IOO54 - :~temal ]~eview Draft:) October I, 1990

TABLE OF CONTENT~; EXECUTIVE SUMMARY ...................................... 4 The Meta-Analvsis Is Fundamentally Flawed And Provides No Ra~_s For Derivin~ ETS ~-,ung Cancer Mortali~v Proiection.~ ................... ae B. C. D. E. F. 12 ~ .................................... 13 Oualitv of Studifi~ ................................. 21 Inclusion of Non-U. S. Studie~ ......................... 30 Misclassification of Smoking Statu.~ ....................... 30 Adiustment for Back~,round Exposu~ ..................... 31 Nicotine sources other than ETS ................ 33 ETS and urinary cotinin_e .................... 33 Urinary cotinine and lun~ cancer ............... 35 Urinary cotinine levels from background exposure, ..... 35 Cultural differences in background exposure ........ 39 0) (iv) (v) (vi) G. The Wilcoxon Analvsi.~ 40 H. ~ ..................................... 42 II. ETS Has Not Been Proven Scientifically To Be A Human Lung Carcinogen .. 43 Statistical Significmlce ............................... 43 Confoundin~ 45 Bias ..... . ..................................... 45 Dose-Re _sponse ................................... 50 Consistency/Breadth of Evider~ ........................ 55 Biologic Plausibility ................................ 57 Weight-of-Evidence ................................ 62 (i) (ii) (iii) Hazard ldentificatigr} ....................... 63 Dose-Res_~nseYExposure Assessment .......... ; .. 66 Risk Characterization ...................... 67 Ao B. C. D. E. F. G. H. EPA's Analysis Of Parental Smokin~ And Childhood Resp_ ir~tory_ Health Superficial. Disguises Policy Preferences As Science And Will Misl~d~ The- Public ............... . .............................. 69

A. The Agency's Analysis is Su~rflci.] ...................... Respiratory Symptoms ...................... Acute Lower Re~iratory Tract llln,~ ............ Pulmonary_ Function ....................... B. The Agency's Review Will Misled_ The Public: ............... M~nstream smoHn~ ~d COPD ................ ETS ~d ~ult ~irato~ h~l~ ................ BIBLIOGRAPHY .......................................... RJR APPEND.IX A: A STATISTICAL REVIEW OF THE EPA REPORT POR APPENDIX B: COMMENTS ON EPA'S HEALTH ASSESSMENT APPENDIX C (DOSIMETRY OF ENVIRONMENTAL TOBACCO SMOKE) RJR APPENDIX C: COMMENTS ON EPA'S HEALTH ASSESSMENT APPENDIX D (ALTERNATE APPROACHF~) '70 72 72 14 75 75 76 77 82 ii

BEFORE THE UNITED STATES ENVIRONMENTAL PROTECTION AGENCY HEALTH EFFECTS OF PASSIVE SMOKING: ASSF__~SMENT OF LUNG CANCER IN ADULTS AND RESPIRATORY DISORDERS IN CHILDREN FR DOC. 90-20013 The Environmental Protection Agency ('EPA" or "the Agency.') released on June 25, 1990, a draft document rifled, Health Effects of Passive Smoking: Assessment of Lun~ Cancer in Adults and Re~_ iratory Disorders in Children, EPA 1600/6-90/006A (IVlay 1990) (External Review Draft) (the "Health Assessment'). Release of the Health Assessment was noticed in the ~.~aJ_Kf, g~ and comment on its technical accuracy and policy implications was solicited. 55 Fed. B_fg, 25874. R.J. Reynolds Tobacco Company ('RJR") submits these comments in response. The Health Assessment concludes that environmental tobacco smoke ('ETS") is causally related to lung cancer in adults and should be designated a Group A carcinogen (known human). The Agency bases this conclusion on what it describes as a weight-of- evidence evaluation of six factors: biological plausibility; broad-based evidence; consistency of response; upward trend in dose-response; detectable association at environmental exposure levels; and effects remaining after adjustment for potential bias. (The Health Assessment at 1-3 to 1-4). A meta-analysis of selected epidemiologic studies of spousal smoking and lung cancer that computes a summary overall relative risk is a crucial element of the Agency's analysis. That summary risk estimate is integrated with assumptions regarding ETS exposure

and the overall incidence of lung cancer in nonsmokers to project the number of annual U. S. lung cancer deaths attributable to ETS exposure. The Agency concludes that the number of deaths attributable annually to ETS exposure lies between 1800 and 6100. The Health Assessment purports to adhere to The Risk Assessment Guidelines of 1986, EPAJ60018- 871045, August 198/(the "Risk Assessment Guidelines') in deriving these projections? The conclusions reached in the Health Assessment are not supported by the scientific data or by the EPA's analysis. The Agency's Health Assessment is inconsistent with its own guidelines for assessing health risks. Weaknesses of the assessment, including uncertainties, assumptions and limitations of the data are ignored. The Agency advances arbitrarily from a statistical exercise to the conclusion that ETS is a group A carcinogen. In general, the Health Assessment is a risk management proposal disguised as a risk assessment and is an abuse of the Agency's discretion. EPA has no statutory authority to regulate indoor air quality. Title IV of Superfund (The Radon Gas and Indoor Air Quality Research Act of 1986) charges EPA with researching indoor air quality issues but expressly withholds regulatory authority. EPA has exceeded its statutory authority and not acted in accordance with law. The Health Assessment constitutes de facto rulemaking. It is not designed to disseminate in a neutral fashion information regarding indoor air quality. Instead, it is clearly d~rected at formulating a basis for regulating smoking in indoor spaces. The Health Assessment is an abuse of the Agency's discretion. PJR's comments address EPA's as~ons, analyses, and cool.ions contained in the Health Asse~ment. No ~_~¢p~nnce by PJR of the EPA's approach in developing the Health Assessment should be inferred.

The Health Assessment also analyzes studies of parental smoking and non-malignant respiratory disease in children. It concludes that there is no proof that the observed associations between parental smoking and various respiratory disorders in children reflect a causal nexus. F_,PA is correct in this regard - there is no scientific proof that the observed associations are causal. However, the Health Assessment is misleading, policy preferences are interwoven with scientific analysis and the data are overinterpreted. This section of the Health Assessment is therefore also arbitrary and capricious. The Agency has the statutory authority to inform the public, not to mislead the public, and doing so is an abuse of its discretion. Our comments are presented in three sections.2 Section I examines EPA's meta- analysis of spousal smoking and lung cancer studies. Section II disproves EPA's identification of ETS as a lung cancer baTnrd. In Section HI, the Agency's analysis of ETS and childhood respiratory disease is reviewed. Thes~ commmts track the format of the Health Asse~meat. "rh¢ body is reslricted to commenti~ ca ]~PA's interpretafi~m of epidemiologic studies of ErS and health. Commeats on Appendices C and D to the Health Assessmeat are appended to these comments, lqo specific comments have been provided for Appendices A and B. ILIR does not, however, concede the accuracy of those appendices. 3

EXECUTIVE S~ARy The Agency has selectively meta-analyzed 22 epidemiologic studies of spousal smoking and lung cancer to calculate a summary relative risk for ETS and lung cancer. In doing so, the Agency has failed to justify the use of meta-analytic techniques and violated three conditions required to achieve a result of any reliability (Glass, e~ a/., 1981; Hunter & Schmidt, 1990). It is apparent that the studies were not evaluated critically before inclusion in the Agency's calculations. One is left to conclude that studies were combined on two bases: lack of exclusion by the National Research Council in its 1986 review of ETS and availability of published data that would fit the chosen statistical technique; no other criteria for inclusion have been identified. The chosen technique is only one of several available and other available techniques,/.e., a weighted general linear model (P.JR Appendix A), are capable of analyzing all published studies. R JR does not agree that meta-analysis is the proper method for analyzing data of the type evaluated by the Agency. The National Research Council has questioned the applicability of any meta-analysis techniques to studies of environmental epidemiology. (NRC 1985, at 218, 219). No meta-analytic technique can replace a reasoned narrative analysis of the study set. At best, if done properly, meta-analysis may provide a means to integrate and summarize overall quantitative trends demonstrated by comparable studies. Combined studies must be comparable in terms of design, conduct and analysis, and must be investigations of the same dependent and independent variables (Glass, eta/:, 1981; Hunter, et aL, 1982; Hedges and Olkir~, 1985; Wolf, 1986; Hunter and Schmidt, 1990). "Comparable" means more than selecting some of the population studies which have 4

evaluated some estimate of exposure to ETS in relation to the incidence of lung cancer in nonsmokers. "Comparable" does not mean that the analyst can, as the Agency has done, simply exclude studies which do not provide similar conclusions, e.g., the Agency has arbitrarily excluded several studies which find no association between spousal smoking and lung cancer in nonsmokers. EPA's justification for combining the case-control studies - testing for statistical homogeneity - is an irrelevant exercise in comparability. Those studies have relatively small sample sizes, and therefore lack statistical power. Not surprisingly, the 95 % confidence intervals overlap. The point estimates, however, vary widely. The combined studies are noncomparable in several respects: 1. The studies were conducted in disparate cultures, ranging from Chinese women to American men and women to Greek women, all of which have differences in lifestyle, other environmental exposures, and perhaps genetic susceptibility. The foreign studies of nonsmoking women drive the summary risk to significance in the technique selected by the Agency. In contrast, if all available data are analyzed, a nonsignificant summary risk estimate for U. S. studies alone of 1.08 (95% CI ffi 0.68, 1.73) is produced, without considering other indicia of appropriateness for inclusion and without adjustment for bias. OUR Appendix A). 2. Combination of case control and cohort studies is inappropriate. The studies have different durations of observation and were conducted at different time periods. 3. Varying histological lung cancer types were evaluated. Some studies (e.g., Brownson, 1987) looked only at adenocarcinoma cases; others (e.g., Trichopoulos, 1981), on . the other hand, specifically excluded all cases of adenocarcinoma.

4. The sources of controls differ greatly, from other cancer patients to atomic bomb survivors to licensed drivers. OUR Table I). Factors used in matching controls vary widely. (PJR Table 11). 5. Though generally a person was considered "exposed" if married to a smoker, the basis for determining whether the sPOuSe was a smoker varied widely. (RJR Table l]I). 6. Inconsistent adjustments were made for confounding factors other than age and marital status, and virtually none of the studies controlled for precisely the same confounding factors. (R JR Table IV). Study selection has exaggerated EPA's summary risk estimate. As a result of EPA's failure to apply rational criteria for the inclusion of studies in the meta~analysis, poor studies have been included and other studies finding relatively low estimates of risk have been -excluded. A notable example of questionable inclusion is the Hirayama study, about which - despite the Agency's contention otherwise -- serious unanswered questions have been raised that cast doubt on the validity of that study's conclusions. An example of studies excluded from the Agency's meta-analysis is the Varela study. This is the largest case control study conducted to date; it examined a U. S. population; and it was excluded by EPA because the statistical method for analysis chosen by EPA could not accommodate its data. The Varela study reported a nonsignificant relative risk of 0.93 (95% CI = 0.55, 1.57). EPA has arbitrarily dismissed the issue of publication bias despite evidence both that researchers are less likely to submit negative results for publication and that journals are less likely to .accept such results for publication. (Chalmers, eta/:, 1990). Failure to verify the existence of non-published studies is a violation of the Agency's duty to ensure that all or a 6

representative sample of all studies of the hypothesis were included in its meta-analysis. Adjustment of the summary relative risk upward for "background exposure" is inappropriate for several reasons. The adjustment assumes that a causal association is demol~su-ated; it is not. The adjustment further assumes that urinary cotinine is a quantitative marker of ETS exposure and that it is related to lung cancer risk; it is neither. Dietary nicotine (present in solanaceous vegetables and other foods) must be controlled for. (Castro and Monji, 1986; Idle, 1990). Moreover, atmospberi¢ nicotine does not appear in constant ratio to other ETS constituents (Nelson 1989, 1990a, 1990b), and its retention by nonsmokers is dissimilar to other ETS constituents. (See Hiller, et a~, 1982; Ingebrethsen, 1989). Urinary cotinine concentrations in nonsmokers are useless for quantifying ETS exposure. In nonsmokers, urinary cotinine is present, if at all, at levels very close to or below current limits of detection (Biber, et a~, 1987), and substances such as caffeine interfere with cotinine determinations (Thuan, et a~, 1989). At best, urinary cotinine is a qualitative marker of atmospheric nicotine exposure. The Agency subjected its overall data to a statistical test of significance using a Wilcoxon analysis. In doing so, the Agency did not apply the same tests for bias (e.g., mis- classification) that it did for its meta-analysis. More important, it applied a one-tall test for significance, when a two-tail test is generally accepted in the scientific literature. This ignores the possibility of values less than one, biases the results, and overstates the observed relative risk. (Denton, 1990). The meta-analysis is fundamentally flawed. Even if it were statistically sound, statistical tests alone cannot establish a causal inference. It is unclear upon what bases EPA

concludes that ETS is a Group A carcinogen. The EPA~s own guidelines (The Risk Assessment Guidelines of 1986, 51 Fed. B_f,g., 33992; Guidelines for Estimating Exposures, 51 Fed. I~g. 34042; Guidelines for the Health Risk Assessment of Chemical Mixtures, 51 Fed. Egg. 34014) require the following: 1. Chance must be ruled out- Only 5 of the 24 individual studies reviewed by EPA reached statistical significance. Without adjustment for bias and confounding, analysis of a variety of subsets (U. S. studies reviewed by El'A; all U. S. studies to date; all studies reported since the 1986 NRC and Surgeon General's reviews; all EPA studies plus 3 recent studies) produces a non-significant summary odds ratio. (RJR Appendix A). These results indicate that chance cannot be ruled out. 2. Possibility of confounding must be ruled out - Confounders and potential lung cancer risk factors are not adjusted for consistently in the studies. Though the Agency summarily concludes that no correlate of-ETS has been identified as an explanation for the observed associations, several researchers have identified lifestyle characteristics reported to be associated with lung cancer risk also associated with ETS exposure. (Friedman, eta/., 1983; Koo, 1989; Sydney, et aL, 1989). EPA has failed in its burden to rule out the possibility of confounding as an explanation for the observed associations. 3. Bias must be excluded as an explanation for any observed association - Smoking stares misclassification and publication bias are not the only forms of bias that receive inadequate analysis by EPA. The Agency apparently concludes, for example, that bias due to a proxy respondent is negligible. Wynder, 1987, however, concluded that U[r]elatives of a nonsmoking lung cancer patient are more likely to report passive inhalation 8

exposure on the part of a relative than relatives of a control patient .... " Therefore, several sources of bias have not been adequately addressed by the Agency. 4. Strength of association must be evaluated - Even for those studies included by the Agency which show a statistically significant association, the association is weak. "Weak" has been characterized as a relative risk less than 2.0 to 3.0. (Wynder, 1987; Doll, 1985). The EPA itself, in 1989, suggested that relative risks less than 5.0 are considered weak. Observed associations in the ETS lung cancer literature are in the range that strains the limits of the epi.demiologic method. The Agency does_not address squarely the problems inherent in interpreting weak associations. 5. A dose-response relationship must be demonstrated - Not one study reviewed by EPA exhibits a statistically significant dose-response relationship when attention is restricted to exposed subjects. Despite the Agency's statement that plots for trend are consistent with a statistically significant association between ETS and lung cancer, it is apparent that the data are consistent with either the presence or absence of a dose-response. (R JR Figure 2). In fact, many of the studies are consistent with an inverse dose-response relationship. 6. Consistency of the association must be evaluated - Inconsistencies in the studies are minimized by EPA. The fact that 80% of the studies are statistically insignificant at the 5 % level does not demonstrate consistency of association. A striking inconsistency is seen when observed associations are broken down by lung cancer histologic type. (See, e.g., Garfmkel, 1981). The two animal inhalation experiments investigating ETS and lung cancer have found 9

no meaningful histopathological differences between animals exposed to EFS and those which were not exposed. (Halcy, 1987a, 198To; Adlkofcr, 1988). One of the most misleading sections of the Health Assessment is that in which it discusses biological plausibility. The Agency relies on the premise that ETS is chemically similar to mainstream tobacco smoke, an unfounded assumption. Even if one were to accept the unproven assumption that smoking is causally related .to lung cance¢, extrapolation from smoking to ETS exposure is unwarranted: 1. ETS is a qualitatively different chemical mixture than mainstream smoke. The biologic effects of one mixture cannot be predicted based on the presence of constituents found in a second complex mixture. (NRC, 1988 at 3). 2. There are many significant differences between ETS exposure and cigarette smoking. (USPHS, 1986 at 7). The Agency has not conducted a weight-of-evidence analysis of ETS and lung cancer. EPA's review is limited to a statistical manipulation of a subset of the epidemiologic data. No evaluations of the physical-chemical properties of ETS, routes and patterns of exposure, structure-activity relationships, pharmacokinetic properties or animal studies is described. The determination that ETS is a Group A (known human) carcinogen is arbitrary and has no substantial evidentiary basis. EPA concludes that no causal association between ETS and childhood respiratory disorders has been established, yet it further concludes that ETS should be treated as a risk factor for acute respiratory diseases and chronic obstructive pulmonary disorders in infants and young children. The Agency has not performed a weight-of-evidence analysis or any l0

other analysis to justify this conclusion. The Agency must quantify the contribution of bias and confounding acknowledged to be present in these studies OVitorsch, 1990) before it can conclude that the weak and inconsistent observed associations reported in some studies indicate a true association between HIS and childhood respiratory disorders. 11

I. The Meta-Analvsis Is Fundamentally Flawed And Provides No Basis For Derivin~ ETS Lun~ Cancer Mortality Projections The t~rm "meta-analysis" was coined to describe statistical methods that were developed by social scientists and others to summarize overall quantitative trends a~'oss studies of a particular topic (Glass, 1976). Meta-analysis was developed to systematically and quantitatively evaluate and integrate results from sets of studies. Ac~:ess to raw data is not required because its results are the unit of statistical analysis. The technique was developed to complement, not supplant, the traditional narrative review and study-I~lying approaches. The Agencymeta-analyzed 22 epidemiologic studie~ of ETS and lung cancer using an extended Mantel-Haenszel procedure. The overall estimate of relative risk for nineteen case- control and throe cohort studies was calculated to be 1.41 (95% CI -- 1.26, 1.57). Adjustment for smoldng stares misclassification reduced the overall relative risk to 1.28 (95% CI -- 1.12, 1.45). EPA's modification for "background exposure" to ETS elevated the summary relative risk to 1.48 (95% CI -- 1.21, 1.87). EPA relies on the metaoanalysis for two purposes: (1) to increase the studies' power to determine whether the observed associations overall are ascribable to chance alone, and (2) to obtain an overall measure of the observed association's magnitude for use in deriving mortality projections. The meta- analysis presented in the Health Assessment is invalid for either purpose. The Agency has not justified its use of meta-analysis, the selection of studies to be combined or the assumptions used in deriving mortality projections. Meta-analytic techniques encompass a variety of statistical methods. Regardless of the method used, at least three conditions must be satisfied for the results to be reliable: 1. Combined studies must be comparable in terms of design, conduct and 12

analysis. They must be investigations of the same dependent and independent variables. Noncomparable studies cannot be aggregated. m Individual studies must be evaluated for quality prior to inclusion in the meta- analysis. Low-quality studies must be excluded. Subject to (1) and (2) above, studies to be combined must represent all or a representative sample of all studies of the hypothesis. (Glass eta/., 1981; Hunter et a~, 1982; Hedges and Olkin, 1985; Wolf, 1986; and Hunter and Schmidt, 1990). The meta-analysis described in the Health Assessment violates all three conditions. Application of meta-analytic techniques to epidemiologic studies raises additional issues. The National Research Council Committee on the Epidemiology of Air Pollution in 1985 warned that the application of meta-analysis to epidemiologic studies must be done cautiously: Although meta-analysis is seductively simple, it contains serious perils when applied to most epidemiologic studies, and its quantitative nature can mask serious flaws in data. In essence, meta-analysis assumes that the results of studies can themselves be treated as random variables with predictable distributions. That assumption might be reasonable for experiments repeated under very similar conditions, but it is rarely so for epidemiologic studies, in which extraneous factors are harder to control and nonrandom errors dominate the random ones. (NRC, 1985, at 218). The Health Assessment contains no discussion of the appropriateness of applying meta-analytic techniques on the ETS and lung cancer literature. The combined studies are not comparable; poor quality studies were included; and, the combined studies are not representative of all studies of spousal smoking and lung cancer. Meta-analysis should not be used unless it can be concluded that the studies of a topic are reasonably comparable (Mann, 1990, at 478). The NRC Committee on the Epidemiology 13

of Air Pollution cautioned that combining noncomparable studies may produce misleading results: The pooling of observations from independent studies to increase sample size is questionable in environmental epidemiology, because it ignores the differences between studies altogether. Numeric combination of results from different studies still has only a small role in epidemiology. (NRC, 1985, at 218, 219). EPA justifies combining the studies by testing for statistical homogeneity of the relative risk estimates presented in the case-control studies.- (The Health Assessment at 3-12). This test does not demonstrate comparability in a relevant sense. Due to the relatively small sample sizes, the 95 % confidence intervals for the relative risk estimates overlap even though the point estimates vary greatly. The absence of power in the individual studies - not similarity between the studies - produces the statistical homogeneity. Statistical similarity among the relative risk estimates does not demonstrate comparability. The Agency notes several differences between the case-control studies but concludes that "[s]tudy differences do not invalidate statistically testing the hypothesis that exposure to ETS is unrelated to lung cancer occurrence." (The Health Assessment at 3-12). No rationale for this statement is offered. The Agency has assumed rather than demonstrated comparability. Demonstrating comparability requires examining the design, conduct and analysis of individual studies. Studies combined by the Agency in the Health Assessment meta-analysis are noncomparable on several bases. The studies draw subjects from disparate cultures. Populations in Greece and Japan differ from U. S. subjects in life style and perhaps even genetic susceptibility. Geographic and ethnic differences could influence host factors and 14

overall health patterns. It is inappropriate to combine studies from disparate cultures with those conducted in the United States. The Agency notes that "a real difference in risk in the populations studied" may account for the different observations produced by the Japanese and American cohort studies by Hirayama and Garfinkel respectively. (The Health Assessment a/ 3-38). At 3-39 of the Health Assessment, EPA suggests that exposure related factors may vary between the two cultures. In fact, the risks observed in the United States studies are statistically different from those observed in the Asian studies. As demonstrated in P.JR Appendix A to these comments, the summary estimate for-risks observed in the U. S. is 1.08 and is not statistically significant (95% CI ffi 0.68, 1.73).3 Differences between the Asian and U. S. studies are unlikely to be attributable to chance alone. Combination of the Asian studies produces a relative risk of 1.37 that is statistically different from the U. S. summary risk (p ffi 0.022).4 This inconsistency undermines EPA's conclusion that ETS exposure causes lung cancer. Instead, it suggests that some life-style factor in the foreign studies is responsible. It demonstrates also that the Asian data should not be used for projecting risk to the U. S. population. The studies also are noncomparable in a variety of methodologic respects. RJ-R's positioa is that a meta-analysis of spousal smoking and lung cancer studies is invalid for the reasons described in these comments. P-JR believes further that the EPA has applied meta-analytic techniques to the spousal smoking and lung cancer literature in a selective manner that is designed to maximize the summary risk estimate. R JR retained George Howard of the Wake Fore~ Unive~ty Bownum Gray School of Medicine to assist it in evaluating ~tatistically the meta-analysis presented in the Health Assessment. His report and a copy of Dr. Howard's curriculum vitae are Appendix A to Note, however, that there is great uncertainty and disagreement even amcmg the Asian studies. Gao a a/., in a study funded and co-authored by the NCI controlled for many confmmders not controlled for in other Asian studies and observed a relative risk of 0.9 (95% CI = 0.6, 1.4) among nonsmoking Chinese women married to smokers. This raises serious questions about the munmary risk achieved by combining other, less well controlled, Asian studies. 15

Combination of case control and cohort studies is inappropriate. When studies with different durations of observation are combined, bias is introduced. If the effect is time dependent, or the number of individuals at risk changes markedly, conventional approaches are likely to produce distorted estimates of the true relative risk. In addition, studies conducted at different time periods may be dissimilar, even though of comparable duration. For example, it is known that levels of air pollution have changed over recent years, presumably affecting any studies of respiratory health. The combined studies evaluated varying histological lung cancer types. For instance, Brownson et al., 1987, restricted their analysis to adenocarcinoma while Trichopoulos, 1981, specifically excluded cases of adenocarcinoma from his study. The Agency has not justified combining studies with such fundamental design characteristic differences. As shown in RYR Table I, the sources of controls in combined studies differ greatly, and include: atomic bomb survivors (Akiba, 1986), bone or colon cancer cases (Brownson, 1987), orthopaedic patients (Chan, 1979, Lam, 1987, Trichopoulos, 1981), colon or rectal cancer patients (Garfinkel, 1985), general population controls (Buffler, 1984, Svensson, 1989), licensed drivers (Varela, 1987), matched neighborhood controls (Wu, 1983), and hospital patients with "non-smoking associated diseases" (Correa, 1983, Kabat, 1984). Moreover, factors used in matching controls in the case control studies varied widely. (RJR Table II). Less than half of the studies match for race, hospital or area of residence. Only two (Geng et aL, 1988; Lee, 1981) matched for marital status. Four (Garfinkel, 1985; Kabat, 1984; Lee, 1986; Varela, 1987) matched for smoking status. The index of exposure varied across the studies in a manner that precludes their 16

Study TABLE I SOURCE OF CONTROLS ISour~ of Co~trol~ Atomic bomb marvivo~ v~th disense~ other ~ c~nce~" BROW Coloa and bone marrow patient~ BUFF Popeh~oa and deu~lent mntrol~ CORR Patients with no--rooking associated diseases GAO Neighborhood GARF(Co) Proqx~tive w.~dy GARF ~ of ~olon or re~mm GENG Not given GILL(Co) Prospective study HIRA(Co) Prospective study HUMB Random telephone sample plus Medicare participants INOU Cerebrovascular decedents KABA Patients with nonsmoking associated diseases KOO District controls LAMT Orthopaedic patients LAMW Neighborhood LEEPatients with nonsmoking usso~iated diseases PERS Women from general population SVEN Population controls TRICH Orthopaedic controls from different hospitals VARE Li~z~xl driven WU I Neighborhood aggregation. (PJR Table III). In general, a woman was considered exposed if her husband was a smoker. However, the basis for determining whether the husband was a smoker varied. In Inoue, 1989, a smoker at home was defined as someone who smoked five or more cigarettes per day. If a husband smoked fewer than four cigarettes per day, his wife was considered unexposed. In contrast, Akiba, 1986, Garfmkel, 1985, Geng, 1988, Humble, 17

TABLE II .. ETS AND LUNG CANCER MATCHING FACTORS USED IN CASE~:ONTROL STUDIES A~'ea or "l"nme or de.h/ Stud~ Sex .A~e Pace SES' residence Hospital in hospital Other ~ • • • • • Vital stares, ~ BROW • • • • BUFF • • • • Vital status CHAN • • • C01L~ • • • • OAO • • • GA_-~F • • • Smoi~g G~TG • • • - M~rital HUMB • • • INOU • • • • KOO • • • Housin~ PERS • • VRal status SVEN • • • VARE • • • Smokin~ st~us 2Participatio~t in program of biennial medical examinations of Radiation Effects Research Foundation 3Matching for smoking and marital status in follow-up study only 1987, Lam, 1987, and Trichopoulos, 1981, used cigarettes per day as the measurement of exposure with zero cigarettes meaning unexposed and one or more cigarettes meaning exposed. Koo, 1987, used cigarettes per day as the measure of exposure with females classified as unexposed if the husband never smoked in her presence and exposed if the 18

Study BROW BUFF CHAN CORR TABLE HI MEASURES OF EXPOSURE Criteria of Exposure Spouse ever smoked cigts or never smoked any Presence of smoker, 4+ vs 0-3 h/d Household member smoked regularly vs no smoker Exposed to tobacco smoke at home or work vs. not exposed Spouse ever smoked cigts or never smoked GAO Lived with smoking husband 20+ vs 0-19 y GARF Spouse or cohabitant ever smoked cigts vs. never smoked any GARF(Co) GENG GmL(Co) Spouse ~urrenfly smoked dgts vs. never smoked any Spouse ever smoked cigts vs. never smoked, any Household ever smoked vs. never smoked i H IRA(Co) Spouse ever smoked dgts vs. never HUMB Spouse ever smoked cigts vs. never smoked any INOU Spouse ever smoked 5 + cigts/d vs. never smoked any KABA KO0 Spouse ever smoked vs. never smoked I Spouse ever smoked in presence vs. never LAMT Spouse ever smoked vs. never smoked any LAMW Spouse smoked in presence 1 y continuously vs. never LEE Spouse ever smoked cigts during marriage vs. never PERS Married to a smoker for at least 2 y vs. not SVEN Exposed to ETS as an adult at home or work vs. never TRICH Spouse ever smoked cigts vs. never VARE Spouse smoked cigts during marriage when living together vs. did not Spouse smoked vs. did not smoke husband ever smoked in her presence. Confounding factors other than age and marital status were adjusted for inconsistently. OUR Table IV). Less than one half of the studies adjusted for the subjects' health and only two adjusted for family medical history. One adjusted for air pollution. Only one third of 19

TABLE IV CONFOUNDING FACTORS CONSIDERED IN STUDIES • • • • • • • • • • • • • • • • • • • • • • • • • • • • • AKIB BROW BUFF CHAN GAO U~(Co) GARP GILL(Co) H~(Co) H~B ~OU ~A K~ P~ SH~U ~CH V~ • • • • • • O. • • -' • • • • • • • • • • • • • • • • • • • • • • its& • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • ~'Scx r~lated factor, e.B., m~nstmal cycle. ~'Mari~al the studies adjusted for socioeconomic class. Eight out of 24 studies reviewed adjusted for diet, potentially the most important risk factor for lung cancer. EPA has not justified combining the studies and has failed to consider the impact of combining noncomparable studies on the Health Assessment's conclusions, 2O

B. Oualitv of Studies Meta-analyses should apply clearly stated criteria for including and excluding studies. Individual studies must be evaluated by these criteria to determine which studies to incorporate. The Agency did not establish criteria for evaluating the studies and provided a summary description for only a subset of 11 of the 23 studies. (The Health Assessment, Appendix A). As a result, poor studies have been included.which have exaggerated the summary risk estimate and other studies have been excluded without justification. The Japane.~e. cohort study reported by Hirayama was included in the meta-analysis despite multiple questions regarding its validity. The Agency concludes at 3-34 of the Health Assessment that Hirayama has "adequately" answered the questions raised about his study prior to 1986. This is not the case. Criticisms which have been levied in the scientific literature include inappropriate statistical analysis (Mantel, 1981; Tsokos, 1981; Rutsch, 1981; Kornegay and Kastenhaum, 1981; Harris and DuMouchel, 1981; Grtmdmann, 1981; Grimmer, 1982; Mantel, 1983), poor control of confounders (Sterling, 1981); bias in the cohort selection (Heller, 1982; MacDonald, 1981), misclassification of lung cancer diagnosis (Grundmann, 1981), exposure estimates (Sutton, 1981; Rutsch, 1982; Schievelbein, 1982), and study design (Lee, 1981; Rutsch, 1982). Recently published reports have identified additional flaws in Hirayama's study. A major problem in assessing the quality of the cohort study is the inadequate reporting of the study design, methodology and results. In particular, Hirayama's published reports fail to describe fully the way in which the study cohort was selected, and give no information about the numbers of persons excluded from the cohort on health grounds or the criteria for such 21

exclusions. No explicit information has been reported on the number of subjects lost to follow-up. When mortality projections, based on Japanese national death rates, were compared with Hirayama's reported cohort mortality, the comparisons revealed external and internal inconsistencies in Hirayama's study (Layard and Viren, 1989). Hirayama's cohort death rates were lower than those for all Japan and the cohort mortality deficits varied greatly among sub-cohorts defined by entry age and marital status. Layard and Viren concluded that the inconsistencies raised the possibility that selection biases in Hirayama's study were distorting the comparisons between exposed and unexposed.groups. They also found that the published data indicate that some 10% of the cohort was lost to follow-up and that comparisons with Japanese national mortality statistics strongly suggest that selection biases exist in the data. These biases invalidate any observed relationship between exposure to ETS -and cancer mortality. CLayard and Viren, 1989). The inconsistent pattern of lung cancer risk seen when the Hirayama data are analyzed by wife's entry age raises questions about the validity of the study conclusions. IGlpatrick found that Hirayama inappropriately adjusted his statistical model of wife's lung cancer risk with husband's age (Kilpatrick, 1988). When wife's age was used to model wife's lung cancer risk, the correlation between wife's lung cancer and husband's smoking lost statistical significance. These relative risks exhibit a strong and highly significant, downward trend with wife's entry age, and there is no elevation of risk at all for women aged 60 years or more at entry. Kilpatrick also has shown that, on the only occasion that Hirayama adjusts female cases by their own ages instead of husbands', the model used by Hirayama is underspecified (Kilpatrick, 1989). Application of a better fitted model results in "husband's 22

smoking" losing its statistical association with wife's incidence of lung cancer. Kilpatrick concluded that "because of this finding, it is important to use the correct analysis of Hirayama's study in further mere-analyses of published studies in order to get a global estimate of the association of spousal smoking with lung cancer death rates in nonsmokers." OrJlpatrick, 1989). Ahlborn and Uberla have criticized Hirayama's study on several bases, including: the study had no preexisting hypothesis regarding ETS and lung cancer, the cohort was not representative of Japan's population, reliance upon death certificates for disease classification, failure to control for confounding factors, smoking status misclassification, and the existence of an age selection bias in the data (Ahlborn and Ubefla, 1988). They have reanalyzed Hirayama's data and concluded that if the age selection bias is removed, a non-significant relative risk results for nonsmoking wives married to smokers. The Agency must evaluate independently whether these criticisms are valid before incorporating Hirayama's data into the meta-analysis without appropriate adjustment of the risk estimates. The Agency should not ignore these issues, especially those to which Hirayama has never responded satisfactorily. Other studies, including that of Trichopoulos, for which the investigators have acknowledged significant shortcomings, should be subjected to similar scrutiny. In contrast, the Agency has excluded arbiWarily from the meta-analysis studies which provide relatively low estimates of risk. Varela's 1987 dissertation is excluded because it fails to provide the "raw data" required for a Mantel-Haenszel estimation. (The Health Assessment at 19). This is clearly an unsatisfactory treatment of Varela's study. The study 23

is by far the largest case-control investigation reported to date and examines a U. S. population. In IUR Appendix A to these comments, Howard uses weighted general linear models to meta-analyze the studies. Inclusion of the Varela study, because of its relatively large size, changes dramatically the summary risk estimation. For U. S. studies only, the summary risk estimation for both sexes becomes RR = 1.68 (9~% CI = 0.68, 1.73). (PJR Appendix A at I0). In view of the change that-inclusion of Varela's data causes, the Agency's determination to employ methods incapable of incorporating Varela's data is highly questionable and a clear abuse of Agency discretion. Similarly, the Agency excluded from the meta-anaiysis Shimizu's (Shimizu, 1988) Japanese case-control study because the raw data were unavailable for deriving a Mantel- Haenszel estimation. As with Varela's data, weighted generalized linear models can accommodate Shimizu's estimates. (PJR Appendix A at I0). The study is essentially null with regard to spousal smoking and contradicts Hirayama's :Iapanese results. A balanced treatment of the overall data requires inclusion of Shimizu's study. By failing to include Varela and Shimizu, the meta-anaiytic result is biased upwards and misrepresents the overall findings in the literature. R~_~ntly, another study investigating spousal smoking and lung cancer in :lapan was reported by Sobue et oZ, 1990. They found no increase in risk in wives married to smoking husbands, RR = 0.94 (95% CI = 0.62, 1.40). This study also should be evaluated by the Agency for inclusion in any future analyses. The Agency's failure to evaluate the quality of individual studies and inclusion of studies identified as being of poor quality has exaggerated EPA's summary risk estimate. Letz¢l et a~, 1988, meta-analyzed I0 case-control and 2 cohort studies of ETS and lung

cancer. A quality rating, based on several factors, was assigned to each study. The sensitivity of the results with regard to different assumptions about effects of smoking status misclassification bias was evaluated. Twenty-seven meta-analyses were performed. Statistically significant results were found only when the analysis was restricted to four studies of low quality. If the lowest quality study, Trichopoulos, was excluded, 95% of the 511 possible combinations of the case-control studies produced nonsignificant results. The quality of the case-control studies greatly influenced the results. For the studies determined by Letzel to be of reasonable quality, a statistically nonsignificant 1111 -- 1.07 was calculated. Combination of the studies stated to be of poor quality produced a RR -- 1.65. The Risk Assessment Guidelines encourage the Agency to "identify the strengths and weaknesses of each assessment by describing uncertainties, assumptions and limitations for each assessment." (51 Fed. ~ 33992). Minimally, this requires EPA to describe the quality of the various studies and the sensitivity of the recta-analysis to inclusion of poor- quality studies. C. EPA should not have conducted a meta-analysis prior to verifying that all or a representative sample of all studies on the topic had been identified. A biased subset of the literature can lead to an unreliable estimate of summary risk. Bias can result if the submission or acceptance of a manuscript is conditioned upon whether the null hypothesis is rejected. Many scientists believe that studies with negative results are much less likely to be submitted for publication (Mann, 1990). Dickersin et a~, 1987, contacted 318 authors of published trials to ask if they had participated in unpublished 25

studies, and "the 156 respondents reported 271 unpublished and 1,041 published trials." Dickersin, 1990, suggests that the direction of the results (positive or negative), and the funding source may affect likelihood of publication. Further, he concludes that large studies, or small studies with large effects may be published preferentially. Chalmers eta/:, 1990, notes that publication bias can occur at three stages. The researcher may fail to publish for various reasons. Pee~" reviewers may be biased, and results rejected by reviewers for one journal may be accepted by another. Finally, post-publication bias may occur in reviews. The impact of bias in research publication is crucial, to resolve in a meta-analysis involving low risk epidemiologic data. For meta-analysis to accurately~ quantitatively summarize a research area, the analysis must include all, or a representative sample of the studies, including unpublished ones. The NRC, 1985, discussed the n_~l__ to ensure inclusion in a meta-analysis of all relevant studies or a subset .representative of the literature: Moreover, for meta-analysis to work, it must be assumed to include all or a representative sample of all studies on a particular question. The possible existence of an unknown number of multiple comparisons, in file drawers or in the minds of researchers, makes this a dubious assumption as well, although procedures have been developed to calculate the number of unretrieved studies that would be required to alter a combined-probability estimate. (2q-RC, 1985, at 218). Begg and Berlin, 1989, found that there is a bias against publishing studies that do not report positive associations, and that this bias can have a "serious impact on meta-analyses." The authors concluded that an important publication bias exists favoring the publication of statistically significant trials. Smith, 1980, reviewed 12 meta-analyses by Glass, Smith and Barton. He found that "[i]n everyone of the 12 instances in which the comparison can be made, the average experimental effect from studies published in journals 26

is larger than the corresponding effect estimated from theses and dissertations. Hunter and Schmidt, 1990, confirmed that effect sizes reported in theses and dissertations differ from values in other publications. Denton, 1990, has shown that author/editor selection biases result in a different population of experiments being "presented to readers than the population observed by experimenters, who generally cull experiments without significant differences before submission to journals, and the population observed by editors, who may conduct similar screenings. This results in an inflated probabifity of type I_error (getting significant results which are spurious), since at a 0.05 significance level, I out of 20 papers (5 %) will on average contain such errors, but if the twenty papers are culled from a set of 100 experiments, 5 of the 20 (25 %) will contain such errors on average. It also biases estimates away from null hypothesis values. Denton concluded that "[o]ne cannot know the filter bias with any accuracy," indicating that it is virtually impossible to make an adjustment for this problem. EPA arbitrarily dismissed the role of post-completion bias. Yet the Agency elsewhere included a discussion of a large case-control study designed to investigate ETS and lung cancer which had not appeared in the published literature at the time the Health Assessment was released (Varela, 1987).s This fact alone should have alerted EPA to the reality of publication bias in the area. The Agency also failed to include in its analysis a thesis reporting the completion of the second American cohort study of ETS and lung cancer Janerich e~ a/., un September 6, 1990, published in The New Eneland Journal of Medicine selected results of the Varela thesis. The relative risk of lung cancer for ever having had a spouse who smoked was 0.93 (95% CI = 0.55-1.57) based on 129 case-control pairs interviewed directly. (Janerich, 199o). 27

(Butler, 1988). That study reports on two cohort samples from a larger cohort of California Seventh Day Adventists. No statistically significant elevation of lung cancer incidence was associated with ETS for 11,060 spouse pairs and 6,467 AHSMOG study subjects. Buffer noted that the study was consistent with there being no ETS effect. The Buffer study is available from Dissertation Abstracts. The Agency's failure to pick up the Butler study in its review is a clear example of post-completion bias. A more subtle form of post-completion bias is the tendency of researchers to gather data on multiple independent variables and report odds ratios only on those for which significant associations are observed. Lloyd et a~, 1986, reported on a case-control study of respiratory cancer in a Scottish industrial community. More controls than cases were found to be ETS exposed at work and at home, but no El'S/lung cancer relative risks were reported. Ires eta/., 1988, reported on a case-control study of lung cancer mortality among women employed in selected industries and occupations in Texas, 1977-1980. No significant increase in odds ratios was obtained for having lived with a smoker but no results were reported. Axelson et a~, 1988, evaluated the possible role of indoor radon for lung cancer risk. According to the investigators, information on the subjects' ETS exposure was obtained. When the data were analyzed with regard to ETS, the Mantel-I-Iaenszel rate ratio was 0.9 for ETS exposure. However, no raw data were provided. The investigators suggested that any future studies of ETS control for indoor radon. At the June 1987 Annual Meeting of the Society of Epidemiologic Research, Reynolds et al., 1987, presented on a prospective study of ETS and cancer in Alameda County, California. Data on ETS and overall cancer and "smoking-related cancer" were presented, however, no data on ETS and 28

lung cancer were presented. Clearly, the data were collected. Moreover, the investigators stated that only the analyses for wives were being reported "beca~use we found no association between passive smoking and cancer among husbands." Alavanja eta/:, 1990, presented at the 1990 Society for Epidemiologic Research Meeting a paper on risk factors for lung cancer among nonsmoking women. Spousal smoking information was obtained and adjusted for in analyses of potential risk factors. However, no data on spousal smoking were presented. At the August, 1990, annual meeting of The International Society for Environmental Epidemiology, He et al., 1990, reported a study of lung cancer in Xuan Wei County, China. In a case-control study of 1 I0 newly-diagnosed lung cancer patients, a significant association was observed for chronic bronchitis and a family history of lung cancer with lung cancer in nonsmoking females. ETS exposure was not associated with lung cancer. None of the foregoing studies were reviewed in the Health Assessment. The preceding are only examples; many additional negative or neutral data sets may exist. The Agency's total treatment of this source of bias consists of a three sentence review of three papers that evaluated the issue. The Health Assessment at 3-33. The Agency appears to have adopted Well's view that publication bias is unlikely to have had any substantial effect on a summary relative risk calculated from the published literature. The Agency's arbitrary conclusion falls short of its duty to provide a discussion of the scientific uncertainty surrounding those assumptions. The Risk Assessment Guidelines. 51 Fed. ~ 33992. If the EPA persists in meta-analyzing the epidemiologic studies, it is incumbent upon the Agency to assess reasonably the impact of post-completion bias on the summary risk by 29

(1) tracking down known data sets that could provide information (Mann, 1990), and (2) adjusting the summary risk for the potential effect of unknown data sets. D. Inclusion of Non-U. S. Studies The Agency incorporates into the meta-analysis both U. S. and non-U. S. studies. Combination of the studies is invalid for any purpose. Lifestyle, ethnic and cultural differences in the study subjects and environments are so significant that the studies are, simply put, of dissimilar independent and dependent variables. Combination. of U. S. and non-U. S. studies to estimate U. S. risk is clearly inappropriate. The Agency at 4-24 of the Health Assessment tacitly acknowledges as much by calculating separately a summary risk estimate for U. S. studies. Inexplicably, the Agency then ignores its own U. S. summary risk estimate, RR = 1.25 (95% CI = 1.03, 1.52), and employs the estimate for all studies, RR = 1.41 (95% CI = 1.26, 1.57), as a departure point for deriving U. S. mortality projections. Even a cursory review of the data suggests that the U. S. risk estimate is different from the overall estimate. For instance, the U. S. risk estimate described in the Health Assessment (1.25) is below the lower 95% confidence boundary (1.26) for the overall risk. When data from the Varela study are added to the U. S. calculation, the summary risk becomes nonsignificant and statistically different from the overa]l risk. (See PJR Appendix A). The Agency is obligated to explain this difference and justify its use of an overall risk estimate for purposes of deriving U. S. mor~lity projections. E. Misclassification of Smoking Status The Agency's adjustment for bias due to smoking status misclassification reduces its 30

reported overall summary risk of 1.41 (95% CI = 1.26, 1.57) to 1.28 (95% CI = 1.12, 1.45). However, the Agency has not addressed adequately the uncertainty accompanying that adjustment. Lee, 1988, has stated that this bias is based on three premises: (1) some smokers are mis~lassified as nonsmokers, (2) smoking is reported as a risk factor for lung eancer, and (3) smokers marry smokers more frequently than nonsmokers do. The size of the adjustment required for this bias depends on the magnitude of any association between smoking and lung cancer, the percentage of smokers misclassified as nonsmokers in the epidemiologic studies, the strength of the smoking marital concordance factor, and the proportion of subjects who smoke. For each of these variables a range of values has been reported. They should be used to create a matrix of the misclassification rates required to explain observed or calculated relative risks. Lee, 1988, reviewed in detail the evidence on each of these factors. Depending on the study designs and populations involved, a range of values was produced. For some populadons, the magnitude of this bias alone is great enough to explain completely the observed associations between ETS and lung cancer. The Agency should assess fully the potential magnitude of this bias for discrete populations during specific time periods. The weaknesses of the data and the range of uncertainty should be described. F. Ad_iustment for Back_~round Ex_vosure The overall summary relative risk reported in the Health Assessment, adjusted for smoking status misclassification, places the excess risk of lung cancer associated with spousal 31

smoking at about 28%. The Agency then adjusts the summary relative risk for "background exposure." That adjustment raises the summary excess risk to about 48% (RR = 1.48, 95% CI = 1.21, 1;8"7). (The Health Assessment at 4-24). The Agency reasons that nonsmokers married to nonsmokers are rarely "truly unexposed" to ETS because low ETS levels may be found in environments outside the home. Thus, the Agency concludes that some part of the baseline lung cancer incidence observed in nonsmokers married to nonsmokers is attributable to ETS exposure from sources other than spousal smoking. To redefine relative risk to be risk relating to zero ETS exposure, instead of to an average background ETS exposure, a second adjustment was made. The Agency adjusts for background exposure based on the ratio of urinary cotinine concentrations in nonsmokers married to nonsmokers relative to the levels observed in nonsmokers married to smokers. The Agency states that the ratio of average cotinine concentrations is three and assumes that "cotinine is a constant multiple of carcinogenic potency of ETS at low doses." (The Health Assessment at 4-28). The adjustment for background exposure is appropriate only if ETS exposure in fact causes lung cancer. The background exposure adjustment provides no evidence that ETS exposure causes lung cancer. Moreover, the Agency's statistical adjustment for background exposure is based on the assumptions that urinary cotinine is a quantitative marker of ETS exposure and is related quantitatively to lung cancer risk. Both assumptions are demonstrably false. An exposure biomarker must meet four criteria to be useful quantitatively: (I) the marker must be specific to the exposure of interest; (2) the marker must provide an integrated measure of exposure; (3) the marker must be a sensitive measure of exposure; and (4) the

marker must relate exposure to disease risk. these criteria. (i) Nicotine sources other than ETS. (Jarvis, 1989). Urinary cotinine meets none of The Agency's assertion that urinary cotinine concentrations in nonsmokers are attributable solely to ETS exposure is false. Sheen, 1988, and Castro and Monji, 1986, have demonstrated that nicotine (a precursor of cotinine) is present in solanaceous vegetables commonly eaten by nonsmokers, including potatoes, tomatoes, eggplant and green pepper. The presence of urinary cotinine in nonsmoking subjects who deny ETS exposure is consistent with the ingestion of dietary nicotine. The dietary contribution of nicotine to the body burden of nicotine and cotinine was estimated by Idle, 1990, to be equivalent to the amount of nicotine absorbed from smoking 1 to 2 cigarettes. Dietary nicotine intake must be considered as a possible explanation why "in several British studies, nearly all nonsmokers had measurable cofinine levels, regardless of reported exposure." (Idle, 1990). The Agency's assertion that "positive cotinine concentrations in 50% to 75%of self-reported nonsmokers, including persons reporting no exposure to tobac.a:~) smoke in the detectable period (up to a few days, depending on the body fluid tested), demonstrate the ubiquity of ETS" fails to consider dietary nicotine contribution. (Health Assessment at 2-2). Dietary nicotine must be controlled for if urinary cotinine is used as a quantitative marker of ETS exposure. (ii) ETS and urine_ cotinine. The Agency assumes that urinary cotinine levels are linearly related to atmospheric ETS exposure. Implicitly, this requires (1) urinary cotinine levels to be linearly correlated with atmospheric ETS nicotine, and (2) atmospheric ETS nicotine to be linearly correlated with ETS levels (including all relevant vapor and particulate 33

phase constituents). These conditions are not met. Nicotine is found predominantly in the vapor phase of the ETS aerosol. (Eudy, 1986). Urinary cotinine concentrations, controlled for diet, reflect only exposure to nicotine - not other vapor phase or particulate phase The peculiar decay behavior of nicotine, which is believed to be caused by its ready adsorption onto and desorption from surfaces, produces measurable nicotine levels in environments where many other ETS constituents have dissipated. (Nelson, 1989, 1990a, 1990b). Nicotine can both disappear rapidly from an environment immediately following smoking or linger in an environment long after other ETS constituents have dissipated. This latter behavior can lead to measurable nicotine in the absence of other ETS constituents. In other words, nicotine exposure can take place in the absence of ETS exposure. Urinary cotinine levels are only loosely correlated with exposure to ETS. The concentration of ETS constituents other than nicotine may decay to background concentrations under environmental conditions that will produce measurable urinary cotinine concentrations. Decay of nicotine in indoor environments is very different from decay of ETS particulate matter and, depending upon the conditions, can either under or overestimate exposure to ETS particles. (Nelson et aZ, 1990b). The ratio of respirable suspended particulate matter to nicotine in indoor environments has been reported to vary from 3.6 to 70 (Rickert, 1988). Recently, deBethizy et ~, 1989, demonstrated that nonsmokers developed measurable urinary cotinine levels following exposure to nicotine in the absence of ETS particulate matter. The retention of nicotine and ETS particles by nonsmokers is also very different. Hiller et al., 1982, reported that about 10% of inhaled ETS panicles are retained. In

contrast, nicotine, at least from mainstream smoke, is retained almost completely. (Ingebrethsen, 1989). ETS nicotine breathed through the nasal passages by a nonsmoker may not even reach the lung due to its high affinity for moist surfaces. (iii) Urinary cotinine and lune cancer. For a biomarker of exposure to provide a valid indication of the health risk of that exposure, the biomarker should be the agent of concern or else covary with the agents of interest so that it can serve as a surrogate measure. To serve as a quantitative marker of risk, urinary cotinine must be the putative carcinogen or a metabolite or bear a linear relationship with the putative _ETS etiologic constituents. Urinary cotinine provides no basis for quantifying lung cancer risk. There is no proof that nicotine or its metabolites are carcinogenic in the lung. The 1988 Surgeon General's Report concludes "that nicotine does not possess direct carcinogenic activity." (USPHS, 1988 at 605). If particulate phase constituents are considered to be the etiologic factors of interest, there is compelling evidence that urinary cotinine is an inappropriate marker for risk adjustment. Thus, urinary cotinine provides no sound basis, empirically or theoretically, for adjusting lung cancer risk. (iv) Urinary eotinine levels from background exposurf, The study by Cummings et at (1990) which has been used by the EPA to conclude that ETS exposure is ubiquitous suffers from serious methodological problems. Cummings et at used the Maehacek and Jiang (1986) HPLC method to screen urine of 663 never and former smokers for nicotine and cotinine. However, the mean urinary cotinine concentrations were unusually high (9.5 ng/ml) for people who claimed no exposure to ETS. One explanation for these high concentrations of cotinine may be the interference of caffeine with the cotinine determination 35

by the HPLC method. Thuan et al. (1989) found that caffeine eluted between cotinine and the phenylimidazole internal standard, causing interference with the cotinine determination. The low cotinine concentrations found in nonsmokers' physiological fluids prevents the use of cotinine as a quantitative biomarker. Those levels are at or near the limit of detection for most methods. (Biber eta/., 1987). There currently exists no standard method for quantifying cotinine at these levels. In lieu of a standardized methodology, individual researchers have developed their own methods based upon immunoassay, gas chromatography or liquid chromatography. Reproducibility of results is also a problem. In a comparative study, Biber et aL, 1987, found significant variability in reported nonsmoker cotinine levels. Measuring urinary cotinine concentrations poses particular problems. Curvall et cd., 1989, found urinary cotinine in two chemically distinct forms: the free base and a glucoside conjugate. Either form can predominate and there are considerable intra- and inter-individual differences. Examples can be seen in R JR Figure 1. Current methods quantify only the free base form. Smoker's Urinary Cotinine 8 Urinar~y excretio~n (_mg/2___4hr) ] ~ FreeBase ~ 4 1 2 3 4 5 6 7 Subject Curvall et al. 1989 The Agency concludes that the ratio Ir~gure 1: Urinary cotinine in smokers: free base and glucoside conjugate. 36

of urinary cofinine for background/spousal ETS exposure is 3.0. Other studies report different ratios. Wald, 1990, has recently published a new study of 184 nonsmokers attending a health screening center in London which gives values for both men and women. That study found that the relative exposure ratios for men, women and both combined are 3.7, 2.1, and 2.6, respectively. Additionally, the ranges show that the relative exposure ratios for men can vary from ratios of 1.2-8.4, for women from 0.9-4.9 and for all from 1.3- :5.2. Noteworthy from this study is the low correlation (0.33) between log urinary cotinine concentration and reported total duration of exposure. The. representativeness of the above data is brought into question due to the fact that the sample of nonsmokers has a population of only 158 nonsmokers' and 26 smokers' spouses from a health screening clinic in urban London. Jarvis et a~, 1984, have presented data for plasma, salivary and urinary cotinine for another London population, a sample of elderly outpatients attending cardiology and vascular TABLE V MEAN VALUES OF COTININE IN NONSMOKERS BY DEGREE OF SELF-REPORTED EXPOSURE Self-Reported ETS Exposure Cotinine, ng ml"l None A little Some A lot Plasma 0.82 1.81 2.52 1.81 Saliva 0.73 2.20 2.80 2.63 Urine 1.55 6.50 8.65 9.36 clinics. These results are shown in IUR Table V. Of course these results are not directly relevant to the ratio between spouses' and their mates' smoking habits, but they do present 37

something of a dilemma. First, there is little concordance among the subjective measures of exposure and cotinine concentration although the results can distinguish among no exposure and the other categories. Second, the ratios among different fluids do not agree. For example, the ratios of "some" to "none" are 3.1, 3.8 and 5.6 for plasma, saliva and urine, respectively. The Agency has presented no rationale for its use of urinary cotinine as opposed to blood or salivary levels. Lee, 1987, found in a survey including 40 different areas in Britain that among nonsmokers married to smokers compared to nonsmokers married to nonsmokers, the ratios of salivary cotinine were for men, 2.9 vs 0.6 and for women 1.0 vs 0.3 ng/ml. These values arc of further interest for two reasons: I) the ratio is greater for men, 4.8, than women, 3.3; yet epidemiologic studies do not indicate an effect for men, and 2) the result found by Lee for women married to a smoker is nearly the same as that found by :larvis for a person reporting no exposure. Another population of 52 British nonsmoldng women in the Birmingham and Fordingbridge areas of England was studied by Proctor eta/., 1990. Salivary cotinine ratios in ngJml for smoking household/working, smoking household/not working, nonsmoking household/working, nonsmoking household/working categories were 1.7, 2.2, 1.3, and I.I, respectively. In U. S. studies the data are sparse as indicated by the few results quoted by the EPA. The study by Coultas et a/., 1987, does not report data directly related to spousal smoldng but does find that "[a]s with children, the prevalence of passive smoking among adults was high but was not strongly determined by the smoking status of other household smokers or by 38

exposure at work. = In a study by Matsunga et aL, 1989, the urinary cotinine concentrations of two matched groups of nonsmokers were compared. Nonsmokers exposed to cigarette smoke for at least 4 hours a day for a minimum of 1 year immediately preceding the study period were compared to a second group of nonsmokers who were minimally exposed to any cigarette smoke; mean cofinirte concentrations of 3.5 and 0.4 ng/ml were found, respectively. This result would provide a differential exposure index of 8.8 for a small but well matched U. S. population. - (v) Cultural differences in background exvosure. The Agency's use of U. S. and European data regarding background exposure for adjusting Asian epidemiologic studies of ETS and lung cancer is internally inconsistent. In reviewing the Hirayama study, the Agency notes that a Japanese wife may spend more time in the home than a wife in a Western culture, and that "background exposure may be lower for Japanese wives." (The Health Assessment at 3-43). Thus, in Japan, there would be a larger relative exposure to EFS between the so-called exposed and unexposed groups. (The Health Assessment at 3-39). The Agency should not then adjust the summary risk (which is driven largely by Asian data) by background exposure estimates from Western cultures. (vi) Conclusion. A number of propositions should be clear:. (I) urinary cotinine is not attributable solely to ETS nicotine; (2) urinary cotinine levels do not correlate linearly with exposure to atmospheric nicotine; (3) urinary cotinine levels have little relation to exposure to ETS constituents other than nicotine; and (4) urinary cotinine levels provide no basis for quantifying lung cancer risks. In short, urinary cotinine is at best a qualitative 39

marker of atmospheric nicotine exposure, but cannot be used as a quantitative measure of ETS exposure. Leaderer, 1990, reported at a workshop on indoor air quality that "[n]icotine and particularly cotinine in physiological fluids are currently the best available biornarkers for ETS but are difficult to relate to actual air exposures." In a review of the literature, Etzel, 1990, of the Centers for Disease Control, U. S. DHHS,concluded that "[t]he literature to date does not support the use of cotinine concentration as a continuous exposure variable in epidemiologic analyses; rather, broad categorizations of cofinine concentrations are more appropriate in view-of the large intraindividual differences-in cotinine concentrations seen with reportedly similar exposures to tobacco smoke." Likewise, Cummings, 1990, concluded that: The relatively modest correlation between reported ETS exposure and urinary cotinine indicates that other factors such as differing metabolic rates and body size may have a confounding effect on the relationship between cotinine levels and questionnaire measures of ETS exposure. In view of this finding, we would recommend against using cotinine levels as a strictly quantitative indicator of ETS. EPA should delete the background adjustment based on urinary cotinine ratios from the Health Assessment. G. The Wilcoxon Analysis The Agency tested the statistical significance of the overall data by a Wilcoxon analysis. A Wilcoxon signed rank analysis of the case-control studies in Table 3.5 found a p- value of 0.0015. The Agency did not evaluate the sensitivity of that analysis to smoking status misclassification (as done for the Agency's Mantel-Haenszel meta-analysis) or for publication bias. ,;0

In KIR Appendix A to these comments, it is shown that if the Wilcoxon analysis is adjusagi for the Agency's assumption that misclassification of smoking status biases the summary risk estimate upward by 14%, the p-value becomes 0.02. While statistically significant, this result is marginal relative to the unadjusted analysis. If an analysis is conducted with different assumptions regarding the magnitude of smoking status misclassification bias, statistical significance is lost. For adjustments above 1.24, the association in Table 3.5 of the Health Assessment becomes statistically nonsignificant. (RIR Appendix A at 6)..A similar analysis was made of the adjusted estiraates from studies listed in Table 3.6 of the Health Assessment. The p-value decreases from highly significant (0.014) to marginally significant (0.04) for the 14% smoking status misclassification adjustment assumed by the Agency. (PJR Appendix A at 6). For misclassification adjustments above 24~ the association becomes statistically insignificant. OUR Appendix A at 6). EPA has not adequately characterized the potential impact of publication bias on its analysis. A further analysis in PJR Appendix A examines the Wilcoxon probability relating spousal smoking to lung cancer as a function of the number of unreported studies and the size of the association observed therein after a 14% adjustment for smoking status misclassification. That exercise demonstrates that the Agency's analysis of Table 3.5 will lose statistical significance if the table fails to include at least three studies reporting an odds ratio of 1.0. (RJR Appendix A at 6.) For Table 3.6, the existence of one study showing no association would cause the result to become statistically insignificant. (R/R Appendix A at 7). Given the clear publication bias in this area, the Agency cannot rely on the Wilcoxon 41

analysis to bolster its case aga/nst ETS. An additional problem related to the analysis reported in Table 3.5 is the use of the one-tail test for s/gnificance. The possibility of values less than one should not be ignored when estimating relative risks. The publication of confidence intervals for relative risks indicates a two-tail approach to estimations of relative risk in the literature in general. Furthermore, Denton, 1990, has shown that the usual screening of test results presented in scientific journals will produce biased estimates of relative risk under the null hypothesis if a one-tail test result is presented while two-tail tests remain unbiased in the presence of such filtering. The bias of the one-tail test will result in the overstatement of relative risk in this case. EPA provides no rationale for its use of a one-tall test and no assessment of the impact of its use on the overall analysis. H. Conclusion ~ Th~ meta-analysis should be deleted from the Health Assessment. Studies of spousal smoking and lung cancer are noncomparable and violate conditions necessary to performing a meta-analysis. EPA has compounded this error by not identifying criteria to use in evaluating the quality of studies prior to inclusion. Poor studies have been included and other studies (with lower risk estimates) arbitrarily excluded. A nonrepresentative subset of all studies of spousal smoking and lung cancer was combined. The Agency's actions amount to an abuse of discretion and suggest that EPA is more interested in building a case for regulating smoking than in assessing ETS and health. 42

II. ETS Has Not Been Proven Scientifically To Be A Human Lung Carcinogen The basis of EPA's conclusion that ETS exposure causes lung cancer is difficult to discern. Based on the meta-analysis of spousal smoking studies,. EPA concludes that ETS is associated with lung cancer but that a causal inference cannot be based on statistical tests alone. (The Health Assessment at 4-1). The Agency then summarily dismisses confounding and sources of systemic bias as explanations for the observed association. EPA's rationale for those conclusions is unclear. Based on those findings and "other evidence," EPA concludes that ETS. is a Group A (known human) carcinogen. (The Health Assessment at 4- 1). Presumably, the other evidence referred to pertains to the factors of: biologic plausibility, consistency of response, upward trend in dose response, detectable association at environmental exposure levels, and broad-based evidence. I-4). (The Health Assessment at 1-3 to In identifying ETS as a lung cancer hazard (concluding that ETS exposure causes lung cancer), the EPA has violated several of the principles set out in its own risk assessment related guidelines. Moreover, the Agency did not analyze the data using criteria typically employed by epidemiologists. (See USPHS 1964; Hill, 1977). These criteria have been developed over the last 50 years to judge whether or not statistical associations derived from observational studies are consistent with causal relationships. Below, epidemiologic studies of ETS and lung cancer are analyzed in terms of the Agency's Risk Assessment Guidelines and additional criteria generally employed by epidemiologists. A. Statistical Sienificance In the EPA Classification System for Cate_~orizin~ Weight of Eviden~ for 43

Carcino~enicitv from Human and Animal Studies, the Agency requires that chance be ruled out before a causal association can be inferred between exposure and cancer in humans based on epidemiologi¢ data. (51 Fed. ~ 33999). The Health Assessment suggests that this criterion is demonstrated because the meta-analytic result remains statistically significant after adjustment for smoking status misclassification bias. That conclusion is based on an arbitrarily selected analysis. The Health Assessment does not convey a full picture of tbe uncertainty, assumptions and limitations of EPA's rationale for that statement. (51 Fed. ~ 33992). _. Less than 20% of all studies to date have found a statistically significant association between spousal smoking and lung cancer. Only 5 of the 24 individual studies that EPA reviewed reached statistical significance. No U. S. study has found a statisticatly significant elevation. Combining the U. S. studies reviewed by EPA in Table 3.5 of the Health Assessment produces a nonsignificant summary odds ratio of 1.28 (95% CI = 0.97, 1.67). (RJR Appendix A at 12). All U. S. studies to date, including Varela, produce an even weaker result (P,R ffi 1.08, 95% CI ffi 0.67, 1.73). (P, JR Appendix A at 10). All studies reported since the 1986 NRC and Surgeon General's Reports combined are weak and not statistically significant (P.,R ffi 1.16, 95% CI -- 0.64, 2.1). (RJR Appendix A at 12). If Varela, Sobue and Shimizu are added to the studies reviewed by EPA, a summary risk estimate consistent with chance alone results (P,R ffi 1.21, 95~ CI = 0.84, 1.75). (RJR Appendix A at 12). Thus, even using the Agency's own analysis, chance cannot be excluded. The Agency's attempt to reach statistical significance by restricting its analysis to a non-representative subset of the existing studies is arbiwaxy and capricious and without

justification. The only relevant subset, U. S. studies, is clearly compatible with chance. Moreover, the analyses presented in RJR Appendix A do not-adjust for bias and confounding. Reasonable assumptions regarding the magnitude of bias and confounding will render the overall summary estimate indistinguishable from 1.0. The Health Assessment in its current form fails to convey how marginal the statistical significance of the data is and how uncertain the assumptions are. In addition to excluding chance, the Agency has identified two other criteria that must be met before a causal association can be inferred between exposure and cancer in humans. 51 Fed. ~ 33999. The first is that "[t]he possibility of confounding has been considered and ruled out as explaining the association." (51 Fed. Kf.g, 33999). EPA states that review and analysis of the epidemiologic studies has not indicated "a correlate of ETS that may explain the observed association between ETS and lung cancer." (The Health Assessment at 4-1). The Agency presents no rationale to support this conclusion. Friedman et aL, 1983, found that many "health-related characteristics" were associated with self-reported hours per week of ETS exposure. They found ETS exposure to be associated with lifestyle characteristics which are also reported to be associated with lung cancer risk, including the increased consumption of alcohol, marijuana, and frequency of hazardous occupational exposures. Friedman et o~ concluded that "[o]ur data also indicate that studies of the effects of passive smoking should consider the correlates of this form of smoke exposure before concluding that it is responsible for some observed effect" (Friedman et al., 1983 at p 404). Sydney et o~I., 1989, found that subjects who reported no ETS

exposure in the home were found to have a higher mean beta-carotene intake than those who reported ETS exposure. Since epidemiologic evidence suggests an inverse relationship between dietary intake or blood levels of beta-carotene and lung cancer, beta-carotene dietary status is likely to be a confounding factor in studies of ETS and lung cancer. Recent reports by Rylander et 02, 1989, and Koo, 1989, have found that other ingested substances, including components of the diet, are risk factors for lung cancer and potential confounders of studies of ETS and lung cancer. Rylander suggested that the absence of control for oral exposures may explain the high incidence of lung cancer in nonsmoklng females observed by Hirayama and that Hirayama's data should be reexamined with respect to such dietary factors. Koo found that data on Chinese women in Hong Kong suggest that consumption of vegetables, meats and fish that are smoked, salted, cured, or pickled, is associated with an increased risk of lung cancer relative to consumption of fresh versions of these foods; and that the former diet was more prevalent among women with smoking husbands than among women with nonsmoklng husbands. Other potential confounders and lung cancer risk factors are adjusted for inconsistently in the spousal smoking studies. Exposures in the workplace have been largely ignored. Several additional lung cancer risk factors for nonsmokers have been identified. (P-JR Table V]). All are potential confounders of the observed association between spousal smoking and lung cancer. They include: dietary cholesterol/fat intake, RR -- 2.2, Goodman et 02, 1988; physical inactivity, RR = 1.6, Albanes et 02, 1989; alcohol consumption, RR = 2:19, Pollack et 02, 1984; genetic and family history, RR = 2.4, Ooi et 02, 1986, RR = 5.3, Samet et 02 1986; the use of certain indoor cooking fuels, RR = 2-3, Gao et 02, 1987, 46

TABLE VI FACTORS REPORTED AS RELATED TO LUNG CANCER RISK F~Aor Family hi~ory of lung cancm" Family histmy of mber¢~o~ ~:amm~vimmin A Samet ~ a/., 1986 Ooi at a/., 1986 Wu at a/., 1988 Salmrai at a/., 1989 Ziegler a a/., 1986 Willett, 1990 Max. RR reported (95~ CO 5.3 (2.2-12.8) 2.4 lO.O (I.I-~0.I) 6.4 2.2 Alcohol intake Pollack at a/., 1984 2.19 (1.3-5.00) Dietary ¢holesteml/fet Goodman a a/., 1988 2.2 (1.3-3.8) Dietary fat intake Wynder at a/., 1987 4-6 Vegetable diet !Jain at aL, 1990 - 0.6 (0.4-0.88) Lack of vegetablm in male diet Le Mtrchand at a/., 1989 2.7 (1.2-6.1) Milk intake Mettlin at aL, 1990 2.1 (1.4-3.2) Drinldng green tea Tewes at aL, 1990 2.7 Hormone therapy in females Adami et al., 1989 1.3 Cooking methods Gao at a/., 1987 2-3 Mumford at al., 1987 Edling et al., 1984 2-3 Kulessa at al., 1989 4.3 (1.7-10.6) 2-3¸ Lees_ at aL, 1987 2.4 (0.8-7.1) Occupation Kvale at al., 1986 2.6 I Motor exhaus~ exposure Hayes at aL, 1989 1.5 (1.2-1.9) Socioeconomic class Brown at aL, 1975 2.6-3.8 Veatilatory function and mucus Lange at a/., 1990 2-4 hypemscretioe Cardiac anomalies Tenkanea at al., 1987 2.4 Physical inactivity Albanes at al., 1989 1.6 (1.2-3.5) Mumford eta/., 1987, and; drinking green tea, RR ffi 2.7, Tewes eta/., 1990. Perusse e~ 02, 1987, have demonstrated familial aggregation in physical fimess, coronary heart disease and pulmonary function measurements. EPA's rejection of confounding factors as an explanation for the observed associations is unjustified. No description of how EPA weighed known and potential confounders is 47

presented. EPA should determine the extent to which the studies are confounded and adjust its analysis accordingly. C. Bias The Risk Assessment Guidelines also require that bias be excluded as an explanation for an observed association before causation can be inferred. (51 Fed. ~ 33999). The Agency's deficiencies in addressing bias due to smoking status misciassification and publication preference have been described above. The Agency has considered inadequately other sources of bias that may explain, at least partially, the observed associations. A particular concern in low-risk epidemiology is the bias inherent in differential reporting of exposures between cases and controls. The Health Assessment at 3-3 concludes that respondent bias can be a source of bias in either direction. However, the patient with lung cancer is likely to have a different attitude than the control patient and their attitude towards questions on ETS exposure levels may lead to over reporting of exposure. The Agency should attempt to quantify the potential range of effects this bias may have on the summary risk estimate. Several of the spousal smoking studies include data on exposure provided by someone other than the case. EPA asserts that information provided by surrogates has been comparable to that provided by the cases. (The Health Assessment at 3-33). This is not evident. Wynder, 1987, concluded that "[r]elatives of a nonsmoking lung cancer patient are more likely to report passive inhalation exposure on the part of a relative than relatives of a control patient ...." Perhaps the most striking evidence of the quantitative effect of respondent bias in studies of ETS and lung cancer is embedded in the data reported by 48

Garfinkel, 1985. Risks varied from 0.77 to 0.83 if husband or wife responded to the questionnaire and increased to 3.57 when a son or daughter responded. A risk of 1.58 for wife was calculated based on non-family members' reporting of husband's smoking. The Agency should assess quantitatively the role that this bias may have played in the stuch'es incorporated in the meta-analysis. D. ~ The strength of an association refers to the magnitude of the observed ri~k~ (The Risk Assessment Guidelines at 1-6). The Risk Assessment Guid_.elines state that the strength of epidemiologic evidence depends, among other things, on the magnitude of the response, 51 Fed. ]~xf~ 33995, and that "epidemiologic studies are inherently capable of detecting only comparatively large increases in the risk of cancer." 51 Fed. ~ 33996. Similarly, the U. S. Surgeon General fists sh"ength of the association as an important criterion to consider. (USPHS 1964). Although there is no precise definition of ~weak," relative risks of less than 2.0 to 3.0 are generally considered to be weak (Wynder, 1987; Doll, 1985). The EPA itself has suggested that any relative risk under 5.0 be considered weak. The Agency organized in 1987 a workshop on the use of human evidence and carcinogen risk assessment. Workshop R _fport on EPA Guidelines for Carcinogen Risk Assessment: Use of Human Evidence, U. S. EPA, September 1989, EPA/625/3-90/017. Groups in that workshop evaluated "straw-man" language proposed by the EPA technical panel related to assessing the need for revision of the risk guidelines. Criteria for evaluating epidcmiologic-based causal interpretations, including strength or magnitude of the association, were considered. After noting that a weak 49

association is more readily explained by bias or confounding than a strong one, the Agency suggested that relative .risk below 5.0 might be considered weak: Should we provide a guideline value, such as a relative risk of 5.0, since magnitude of association can also depend on, for instance, the variety and range of magnitudes of exposure present and the rarity of the cancer7 Workshop R _eport on EPA Guidelines for Carcinogen Risk ~,~sment at 12. By this guideline, all of the risks reported in the epidemiologic studies of ETS and lung cancer clearly are weak. Of the 25 sex-specific relative risks reported in Table 3.5 of the Health Assessment, none are above 3.0 and only five are above 2.0. Five are at 1.0 or below. All estimates of the summary risk are below 1.5. The observed associations are in the range that strains the limits of the epidemiologic method. Weak associations must be interpreted with extreme care as they are even more likely to be artifactual or spurious than strong ones. The Health Assessment does not address squarely the problems inherent in interpreting weak association, in violation of EPA's Risk Assessment Guidelines that require identification of uncertainties, assumptions and limitations of a risk assessment. (51 Fed. KC~ 33992). E. Dose-Re _sponse The absence of a dose-response relationship decreases the likelihood that an observed association is causal. (The Risk Assessment Guidelines at 1-11; 51 Fed. ~ 33999). The Health Assessment at 1-4 states that the epidemiologic data exhibit an upward trend in dose response. However, the Agency's analysis in Section 3.4 of the Health Assessment concerns a test for trend for exposure effect - not dose-response. The Agency statement that the plots for trend are consistent with a statistically significant association between ETS and lung cancer is misleading. When error bars are superimposed on the bar graphs in Figure 3-4 of 50

AKIB (c~g./day) 2.5 ........." .......................................... 1o5 ....................... 0°5 ........................ 0 20-29 30 1-19 GARF (cig./day) 2o 1-9 10-19 >_20 HUMB (clg.lday) 4' I ............. ÷ ................................... i ............. 4 1-20 .)21 2.5' 2' 1.5" 1" 0.5" 0 GAO (tot. yrs.) 20-29 30-39 GENG (cig./day) 40 1-9 10-19 .~20 INOU (¢ig./day) 5-19 ~20 Figure 2: Rcpormd dose-response relationships with error bars. 5!

il

the Health Assessment, it is apparent that the data are consistent with e/ther the presence or absence of a dose-response relationship. (See PJR Figure 2). The Agency applies a similar analysis to conclude that-Garfinkel's cohort data are consistent with either the presence or absence of a dose response. (The Health Assessment at 3-35). None of the 23 studies reviewed in the Health Assessment exhibit a statistically significant dose-response relationship when attention is restricted to exposed subjects. (Layard, 1989). Nor do the data taken as a whole. A meta-analysis of dose-response data presented in the Health Assessment is developed in IUR Appendix A to these comments. Each study was scored for the strength of dose-response observed. The scores from the individual studies were pooled and a test statistic created. The distribution of the test scores under the null hypothesis were estimated by Monte-Carlo methods. When the tests were restricted to exposure levels (the no-exposure levels were excluded) the summary dose- response statistic was insignificantly different from the null hypothesis. 16). (R.IR Appendix A at Finally, the EPA has presented a biased subset of the measures of dose-response that have been reported. In addition to the data cited by the Agency, Aldha, 1986, found that the relative risk of women exposed to husbands' cigarette smoke for 20 or more years was less than the relative risk for women exposed to husbands' cigarette smoke for fewer than 20 years. He found that wife's lung cancer incidence decreased with husband's duration of smoking cigarettes while married. For husband's smoking 1-19, 20-39, and 40+ years, the observed odds ratios were 2.1, 1.5 and 1.3. Similarly, in Garfinkel's 1985 study, he found that the women exposed during the last 53

five years had an OR (adjusted for hours of exposure per day) of 1.28 (95 • CI = 0.98, 1.66) and those exposed for the last 25 years had an OR of 1.12 (95~ CI = 0.81, 1.42). No increasing trend with increasing exposure was apparent in either group. In the five-year exposure group, the OR went down with increased exposure, and the OR in each of the exposure groups was not statistically significant. Women reporting 1-2, 3-6 and greater than 7 hours of exposure per day to the smoke of others within the last five years exhibited odds ratios of 1.59, 1.39 and 0.94 for ETS exposure. In 1983, Koo reported a higher odds ratio for lung cancer in nonsmoking wives when husbands smoked less rather than more. In a retrospective study of 200 female lung cancer cases among Hong Kong Chinese women, data were collected on the lifetime estimated hours of exposure to ETS from the home plus work environments, years of such exposure, and the daily number of cigarettes smoked by the husband. Only for husbands who smoked fewer than 15 cigarettes per day was wife's relative risk statistically significant. No dose-response relationship among any of the measurements of ETS was statistic.ally significant. Koo in a 1984 expansion of her 1983 study, reported higher relative risks for wives with less than 35,000 hours of exposure (RR -- 1.28) than for wives with greater than 35,000 hours (RR -- 1.02) Similarly, Humble reported that "It]he exposure response pattern for cigarettes smoked daffy showed higher odds ratios for subjects whose spouses smoked a pack or less per day (OR -- 2.8) than those whose spouses smoked greater amounts(OR -- 2.2)." Koo in 1987 attempted to obtain a more accurate reflection of lifetime EFS exposure by evaluating the husband's smoking habits, lifetime estimate of total years of exposure, total hours of exposure, mean hours per day of exposure, and total cigarel~ ~ day smoked by 54

each household smoker and found no dose-response result for any exposure measurement. Except for total years, all of the relative risks for the high levels of exposure were below those for the low or intermediate levels for each form of measurement. For instance, wives whose husbands smoked more than 21 cigarettes per day exhibited a relative risk of 1.19 while wives of husband's smoking 1-10 cigarettes per day had a relative risk of 2.33. Neither result was significant. Similarly, women exposed to HIS for 20-34 years had a relative risk of 1.36 while women exposed for 1-19 years had a relative risk of 1.95; women reporting a total exposure of greater than 20,000 hours had a lower relative risk (1.42) than women reporting less than 10,000 hours of exposure (1.68); and, women being exposed to the smoke of more than 21 cigarettes per day exhibited a relative risk of 1.21 while women reporting exposure to fewer than 10 cigarettes per day exhibited a relative risk of 1.83; finally, women reporting exposure for greater than 1.5 hours per day had lower relative risks than women reporting exposure for less than 1.5 hours. F. Consistency/Breadth of Evi~len~ The Health Assessment states that the epidemiologic studies provide consistent support for an ETS/lung cancer link because 18 of the 24 studies reviewed found a relative risk greater than 1.0. EPA then suggests that the studies provide broad-based evidence because they were conducted on subjects from eight different countries by different teams of investigators employing different study protocols. Both analyses are overly simplistic and fail to convey the limitations and weaknesses of the data and the assumptions inherent in the Agency's rationale. (51 Fed. Ke~ 33992). Most formulations of criteria for use in evaluating epidemiologic studies for evidence 55

of causality list ..consistency of the association" as a criterion. (See, e.8., USPHS 1964). Finding the same association in several different studies on cliffer'~t subjec~ provides evidence that an observed association is not an artifact specific to a particular group of subjects. In this sense, consistency across studies is reassuring. Consistency does not, however, assure the validity of the observations. All of the studies of a particular topic can suffer from systematic sources of bias or confounding that spuriously elevate the observed associations. Thus, conclusions regarding consistency are illusory if systematic sources of bias and confounding are responsible for the observed associations. A primary weakness of the Health Assessment is the Agency's failure to evaluate systematically each study for evidence of bias and confounding. A set of criteria should be applied to each study to determine whether it was validly designed, conducted and analyzed. Poor studies should be rejected at this stage. Remaining studies should then be individually adjusted for bias and confounding. :ludgments regarding consistency should be made based on the results. In its current form, the Health Assessment obscures with statistical analysis the uncertainty and weaknesses of the underlying data and reaches a meaningless judgment regarding consistency. The Health Assessment does not describe many inconsistencies in the studies. At 1-4, the Health Assessment states that evaluation of the total study evidence from several perspectives leads to the conclusion that the observed association between ETS exposure and lung cancer incidence is not attributable to chance. However, 18 of the 24 studies reviewed by EPA axe statistically insignificant at the 5 % level. No study of a U. S. population is significant at that level and only one study of a Western population (Greece) shows statistical 56

significance. EPA has ignored the striking inconsistencies seen when the observed associations are broken down by lung cancer histologic type. Lung cancer is not a single disease but may encompass several morphologically and clinically distinct diseases. There appears to be a general consensus in the literature that there are distinct distributions of histological patterns expressed for specific causative factors. If the reported association between EI'S and lung cancer is real, consistent histological patterns should be seen across different studies. However, marked differences exist with respect to type of-lung cancer observed in studies presenting histologic specific analyses. For instance, in his case-control study, Garfinkel, 1981, found "no consistent pattern between husband's smoking and wife's lung cancer type." G. Biologic Plausibili~ The Health Assessment concludes that it is biologically plausible that ETS causes lung cancer because: ETS is token up by the lungs and distributed throughout the body. The similarity of carcinogens identified in sidestream and mainstream smoke along with the established causal relationship between lung cancer and smoking make it reasonable to suspect that ETS is also a lung carcinogen. (The Health Assessment at 1-3). No further analysis is presented. Of all the statements made in the Health Assessment, the preceding is among the most misleading. EPA fails to weigh relevant biologic information, including that there are no animal studies - a requirement for drawing a causal inference - confirming the observed associations. No notice is taken of the uncertainties inherent in predicting the toxicologic profile of one complex mixture based on its chemical similarity to a second. 57

Only two animal inhalation experiments investigating ETS and lung cancer have been reported. Both studies found no meaningful histopathological differences between animals exposed to ETS and those which were not exposed. In a study conducted by the American Health Foundation, (Haley, 1987a, 1987b) the investigators exposed one group of hamsters to mainstream smoke and another group to ETS. Animals exposed to mainstream smoke and ETS lived longer than the sham treated controls. The investigators reported that overall there was no marked increase in tumor incidence in animals exposed to either mainstream smoke or ETS after 18 months of exposure. The second study was a 90-day ETS inhalation study of rots and hamsters (Adlkofer, 1988). Animals were exposed to ETS concentrations 100 times those concentrations encountered by nonsmokers. These researchers reported no histopathological differences between exposed and control animals. Electron microscopy revealed pulmonary changes which could be expected to occur under similar exposure conditions with other sources of respirable particles. In the absence of valid experimental and mechanistic evidence, EPA attempts to rely on mainstream smoking data to demonstrate the plausibility that exposure to ETS causes lung cancer because of the presence of certain chemicals in both mixtures. That reliance is premised on four unproven assumptions: (1) smoking is a causative factor for lung cancer in humans; (2) mainstream and sidestream smoke are qualitatively similar mixtures regarding the presence of putative etiologic constituents; (3) ETS exposure is a quantitative variant of smoking and will produce the same diseases as smoking is alleged to produce; and (4) that there is no safe level of ETS exposure. Smoking and lung cancer are poorly understood. Research conducted over several 58

decades has not proven that smoking causes lung cancer. Moreover, ETS exposure cannot be treated as a quantitative variant of smoking for projecting statistical risk. ETS is a qualitatively different chemical mixture than mainstream smoke and exposure to ETS differs physiologically from smoking. Scientists in the fields of chemistry, toxicology, car~-'inogenesis, and epidemiology agree that it is not possible to identify any particular smoke constituent(s) as the alleged cause of disease. [-u']nderstanding [of the] components which ]nay act on the cardiovascular and respiratory system is embarrassingly lacking, if not outright contradictory .... Uncertainty about the specific attribution of risk to individual smoke components may be greater than ever now .... (Gori, Banbury Report, 1980 at 355.) Richard Peto reviewed 30 years of research and concluded that no particular constituent of mainstream smoke has been demonstrated to be carcinogenic to smokers (Hoffmann, Banbury Report 1982.) Moreover, knowledge of the chemical constituents of a mixture as complex as smoke is of questionable utility in predicting the mixture's biological properties: Cigarette smoke contains over 3,800 compounds... The biologic effects of this galaxy of compounds are virtually impossible to predict from a knowledge of individual constituents, not only because of their huge number, but also because transient chemicals generated during the process of pyrolysis, oxidation, and free radical formation might dissipate or change with time and temperature. NRC, 1988 at 28. The biological effects of one mixture, such as ETS, cannot always be predicted based on the presence of constituents also found in a second complex mixture such as mainstream cigarette smoke. As noted by the National Research Council: 59

Consistency of analytic results between mixtures implies an increase in the predictability of their effects. The more similar two mixtures are chemically, the more similar their toxicologic properties are expected to be. However, even mixtures that are relatively well characterized sometimes have unexpected toxic effects. One must be prepared to look for unexpected results of exposure to complex mixtures, because of the potential disjunction between chemical analysis and biologic effect, complex chemical mixtures are more likely to produce unexpected results than are individual chemical substances, for several reasons. Mixtures are composed of various substances, exposure to which can be expected to be associated with different toxicities. The constituents of a mixture sometimes combine chemically to produce new compounds with different toxicities. The presence of some materials might mask, dilute, or increase the toxicity of other materials. Such phenomena, referred to as interactions, can amplify or reduce anticipated effects. Moreover, different doses of separate materials might increase the bioavailability of materials that are otherwise nontoxic at the doses present in the mixture. (NRC, 1988 at 3). Caution should be exercised when attempting to predict the biological • effects of one mixture based on the composition of another, even when the mixtures are very similar and well characterized. Mainstream smoke and ETS are not very similar. Although mainstream smoke has been extensively characterized. ETS is not well characterized.~ The Surgeon General, 1986, concluded that the knowledge of cigarette smoke chemical composition is of limited assistance for predicting ETS-exposure effects: Comparison of the relative concentrations of various components of SS and MS smoke provides limited insights concerning the toxicological potential of ETS in comparison with active smoking. As described above, SS characteristics, as measured in a chamber, do not represent those of ETS, as inhaled by the non-smoker under nonexperimental conditions. Further, the dose-response relationships between specific tobacco smoke components and specific diseases are not sufficiently established for the necessary extrapolations from active smoking to environmental tobacco smoke exposure for individual agents.

(USPHS, 1986 at 24). In short, hazard identification based on the presence in ETS of putative carcinogens found in tobacco smoke does not establish causation, it presumes causation. A second major flaw is that such reliance ignores the many differences between ETS exposure and cigarette smoking. These differences are well described in the literature. The Surgeon General has acknowledged that exposure to each mixture is fundamentally different: Mhinstream cigarette smoke is the smoke drawn through the tobacco into the smoker's mouth. Sidestream smoke is the smoke emitted by the burning tobacco between puffs. Environmental tobacco smoke results from the combination of sidestream smoke and the fraction of exhaled mainstrem~ smoke not retained by the smoker. In contrast with mainstream smoke, ETS is diluted into a larger volume of air, and it ages prior to inhalation. (USPHS 1986 at 7). Blot andFraumeni, 1986, note that "[c]ontributing to the difficulty of assessing exposure to lung tissue are different routes of exposure, with passive smokers generally inhaling through nasal passages, as opposed to the mouth and throat among direct smokers. The Surgeon General, 1986, has also noted the physiological dissimilarity between cigarette smoking and ETS exposure: The breathing patterns for the inhalation of MS and ETS are also different: MS is inhaled intermittently by the smoker with an intense inhalation, often followed by a breathhold that results in a more equal distribution. Environmental tobacco smoke on the other hand, is inhaled continuously with ~iclal breaths when the passive smoker is at rest and with deeper inhalations when the passive smoker is physically active. Breathholding does not normally occur with tidal breathing. (USPHS, 1986, at 25). 61

The retention pattern of ETS particles also differs radically from the pattern seen for cigarette smoking. (The Health Assessment at C-28). Fo~ this and other reasons, the NRC concludes that data on cigare~ smoking do not provide a basis for predicting ETS exposure effects: Because the physicochemical nature of ETS, MS, and SS differ, the extrapolation of health effects from studies of MS or of active smoke~ to nonsmokers exposed to ETS may not be appropriate .... Laboratory studies in conjunction with epidemiologic investigations are needed to clarify possible health effects of exposure to ETS in nonsmokers. (NRC, 1986 at 8). H. Weight-of-Evidence Carcinogenic risk assessment is a process which generally requires the following steps: Hazard Identification - the determination of whether a substance is or is not causally linked to cancer;, Dose-Re _sponse Assessment - the determination of the relationship between the magnitude of exposure and the probability of occurrence of cancer; Exvosure Assessrrlf0t - the determination of the extent of human exposure, and; Risk Characterization - the description of the nature and the magnitude of human risk including uncertainty. (NRC, 1983). EPA's descriptions of the risk assessment process suggest that any complete risk assessment must contain these four components. In addition to the Risk Assessment Guidelines, EPA published in 1986 several other sets of guidelines to encourage consistent practices in the preparation of risk assessments, including: Guidelines for Estimating 62

Exposures, 51 Fed. R~. 34042; and Guidelines for the Health Risk Assessment of Chemical Mixtures, 51 Fed. ]~g. 34014 (September 24, 1986). EPA's guidelines were published "to guide Agency evaluation of suspect carcinogens in line with the policies and procedures established in the statutes administered by the EPA." (51 Fed. ]~.g. 33992). They were intended to ensure the review of the full body of scientific information on each substance and the use of the "most scientifically appropriate interpretation to assess risk." Agency scientists are encouraged to "identify the sU'engths and weaknesses of each assessment by describing uncertainties, assumptions, and limitations as well as the scientific basis and rationale for each assessment." 51 Fed. ~. 33~)'2. In drafting the Health Assessment, the Agency has failed to adhere to important elements of its own guidelines. (i) Hazard Identification. EPA labels the process of drawing a causal inference for risk assessment purposes as "hazard identification." (The Risk Guidelines, 51 Fed. ~ 33~)93). The Risk Assessment Guidelines recommend that the hazard identification section of a risk assessment "contain a review of the relevant biological and chemical information bearing on whether or not an agent may pose a carcinogenic hazard." (51 Fed. ~ 33~94). To the extent that the data are available, the hazard identification section should contain information on the physical and chemical properties of the chemical(s) under study; mutes and patterns of exposure; structure-activity relationships; metabolic and pharmacokinetic properties; toxicologic effects; short-term tests; and long-term animal studies. Cl'he Risk Guidelines, 51 Fed. ~g~ 33~4). The Health Assessment does not follow the framework set out in the Risk Assessment Guidelines for rnalcing weight-of-evidence judgments regarding hazard identification. The 63

Health Assessment fails to describe either the weakness or absence of data on most of these factors. The Health Assessment is, essentially, a statistical analysis of some, but not all, epidemiologic studies of spousal smoking and lung cancer and a theoretical discussion of alternative dose-response models. It contains no discussion of the physical and chemical properties of ETS nor a comparison of ETS and mainstream smoke, despite basing its conclusion that a b~,~rd is biologically plausible on similarities between ETS and mainstream smoke. It does not contrast physiologic differences between smoking and EFS exposure. It contains no summary of metabolic or pharmacokinetic information as the Risk Guidelines recommend. Negative experimental data and the absence of mechanistic data are not discussed. (51 Fed. Rf.g. 33994). The Health Assessment does not discuss the numerous deficiencies and limitations of the Agency's hazard identification. In addition, the Agency has selectively included information that suggests that ETS may be a hazard while excluding information to the contrary. The Agency's failure to identify and explain these omissions suggests a biased presentation more akin to strength-of-evidence analysis than a weight-of-evidence analysis. EPA's statistical analysis does not provide a true evaluation of the strengths and weaknesses of the studies as the Risk Assessment Guidelines require. The Risk Assessment Guidelines acknowledge criteria for evaluating the adequacy of epidemiologic studies: They include factors such as the proper selection and characterization of exposed and controlled groups, the adequacy of duration and quality of follow-up, the proper identification and characterization of confounding factors and bias, the appropriate consideration of latency effects, the valid ascertainment of the causes of morbidity and death, and the ability to detect specific effects. Where it can be calculated, the statistical power to detect an appropriate outcome should be

included in the Assessment. (The Risk Assessment Guidelines; 51 Fed. ~ 33995). The Agency reviews individually only 11 of the 23 studies and does not discuss those in detail. The Health Assessment contains a limited discussion of potential sources of bias, primarily focusing on smoking status misclassification. The Agency does note that in 9 of the 11 studies reviewed, the researchers failed to adjust for confounding, but implicitly concludes that there is no need to adjust quantitatively the meta-anaiysis. The Health Assessment contains no meaningful discussion of quantitative adjustments required for mischaracterization of exposed and control g~ups or the failure to ascertain validly causes of morbidity and death in the studies. The Agency performs the meta-analysis primarily to overcome the lack of statistical power in the individual studies. However, when all available data are included, the combined studies fail to exhibit statistical significance. (RJR Appendix. A at 12). The Agency should determine the power of those studies combined to detect an increased risk and explain their failure to do so. The Risk Assessment Guidelines provide that three criteria must be met before a causal association can be inferred between exposure and cancer in humans based on epidemlologic evidence: (1) that there be no identified bias that could explain the association, (2) that the possibility of confounding be considered and ruled out as explaining the association, and (3) that the association not be due to chance. (The Risk Guidelines, 51 Fed. Ke~ 33999). The Health Assessment fails to apply the EPA's own criteria to the studies individually. Much of the Health Assessment is a contrived effort to avoid doing so. In fact, a review of the individual studies using these criteria reveals that the studies are so 65

weak, confounded and biased that no conclusions can be drawn. The l~islc Assessment Guidelines provide also that confidence in inferring a casual association is increased when the observed associations are strong, when thereis a dose-response relationship, or when a reduction in exposure is followed by a reduction in the incidence of cancer. (51 Fed. Brig. 33999). The Health Assessment ignores the relative weakness or small .magnitude of the observed association for spousal smoldng and lung cancer. Because no dose-response relationship is observed, EPA attempts to rely on a test for trend. However, even utilizing the Agency's own analysis, the data are consistent with either the presence or absence of a trend. Finally, the Agency fails to note that there have been no relevant intervention studies. Ultimately, the Agency's hazard identification is based on a purely statistical analysis. The conclusion that ETS causes cancer is based upon a combination of questionable assumptions and unfounded conclusory statements and fails to follow the Agency's Risk Assessment Guidelines. No weight-of-evidence evaluation is made. (ii) Dose-Re _sponse/Exposure Assessment. EPA has issued separate guidelines for exposure assessment. The Agency has identified the need for consistency with respect to its assumptions about typical exposure situations and with respect to the characterization of uncertainty of exposure estimates. (51 Fed. Ke~g. 34043). The EPA guidelines provide that: It is important to convey an appreciation of the impact of the strength and weaknesses of exposure assessment on the overall cancer risk assessment process. The assumptions, approximations, and uncertainties need to be clearly stated because, in some instances, these will have a major effect on the risk assessment.

An attempt should be made to assess the level of uncertainty associated with the exposure assessment which is to be used in a cancer risk assessment. This measure of uncertainty should be included in the risk characterization.., in order to provide the decision r~l~_,~" with a clear understanding of the impact of this uncertainty on any final quantitative risk estimate. (51 Fed. B.~g. 33998). The Health Assessment omits completely any discussion of the assessment of ETS population exposure. The Agency provides no integrated exposure analysis combining the ~ of the exposed populations, the duration, frequency and intensity of exposure as the exposure guidelines require. The only exposure discussion is contained in the Agency's evaluation of background ET$ exposure and is based purely on urinary cot/nine levels. The lack of any exposure assessment renders the epidemiology based dose-response assessment groundless. (iii) Risk CharacterizatiOn. The EPA Risk Assessment Guidelines require that uncertainties and assumptions in a Risk Assessment be presented: Whichever method of presentation is chosen, it is critical that the numerical estimates not be allowed to stand alone, separated from the various assumptions and uncertainties upon which they are based. The risk characterization should contain a discussion and interpretation of the numerical estimates that affords the risk manager some insight and degree to which the quantitative estimates are likely to reflect the true magnitude of human risk, which generally cannot be known with the degree of quantitative accuracy reflected in the numerical estimates. (51 F~. ~¢.~. 33999). As pointed out many times in these comments, the Health Assessment is woefully deficient in identifying for the public assumptions and sources of uncertainty. A prime example is the Agency's assertion at 4-32 that "[i]t is unlikely that the true number of LCDs per year in U. S. nonsmokers lies outside the interval defined by the two extreme 67

values, 1800 and 6100." This statement is trans-scicntific and fails to adhere to the Agcncy's own guidelines. It reflects the EPA's antismoking policy orientation - not the scientific data.

lrl. EPA's Analysis Of Parental Smoking And Childhood Respiratory Health Is Su~rficial. Disguises Policy Preferences As Science And Will Mislead The Public. A. The A~,encv's Analysis is Su~e~cial The Health Assessment states that the overall weight-of-evidence suggests that parental smoking is associated with an increased incidence of respiratory disorders in children. The general approach taken by the Agency is to supplement the conclusions found in the 1986 reports of the U. S. Surgeon General and National Research Council with a cursory, narrative review of subsequently published epidemiologic studies. No original analysis of studies reported in the 1986 review literature is undertaken and no evaluation of the appropriateness of the conclusions contained therein is made. EPA's own review is limited to the epidemiologic literature. No weight-of-evidence analysis as contemplated in the Agency's own risk guidelines is performed,e The Health Assessment contains no discussion of: structure-activity relationships, metabolic and pharmacokinetic properties; mechanistic aspects of respiratory disorders; or animal data. (The Risk Assessment Guidelines; 51 Fed. Reg. 33994). EPA's review of the epidemiologic data is trivial. The Agency describes several sources of bias and confounding as potentially responsible for the observed associations, including: parental respiratory disorders, parental recall bias, domestic exposures to heating and cooking fuels, genetic-based predisposition and familial history of respiratory disorders, and socioeconomic status or class. However, with no quantitative analysis, the Agency arbitrarily concludes that it is improbable that bias and confounding could account for the EPA's Risk Assessment Guidelines were developed to assist Agency personnel with the carcino~c risk assessment process. However, most of the principles are equally relevant to assessing non- malignant health risks. 69

observed associations. No rationale for the conclusion is described. More is required of the Agency than conclusory statements. The Agency states further that the epidemiologic studies have controlled for bias and potential confounders to the extent possible. Clearly, this is not the case. Significant potential sources of bias and confounding are not controlled for in any individual study. The Agency should assess quantitatively the likely magnitude of confounding in each study individually. The role of bias should be analyzed similarly. EPA categorizes the epidemiologic studies of ETS and childhood respiratory health into four groups: (1) respiratory symptoms; (2) acute respiratory illness; (3) pulmonary function, and; (4) related results including middle ear effusion and asthma. The Agency did not analyze the epidemiologic data using such standard criteria as strength, statistical significance, specificity, dose-response, and consistency. (Hill, 1977; USPHS, 1964). The result from each group of studies is similar. The data are weak, inconsistent, and largely statistically nonsignificant. Bias and confounding are controlled poorly, if at all. Minimal evidence of dose-response is observed. The associations are nonspecific and not confirmed by animal studies. The studies provide no basis for concluding that an association has been proven. Clearly, no evidence of causation is presented. Additional comments follow on EPA's review of each study set. (i) Re~iratory Symptoms. EPA concludes that there is no apparent single source of systematic bias that might explain the associations observed between parental smoking and respiratory systems (cough, sputum and wheezing) in children. EPA then cites supporting factors for its conclusion that parental smoking increases the incidence of respiratory symptoms, including: (1) evidence that maternal smoking tends to be more of a factor than 7O

paternal smoking in the first one or two years of life; (2) an observed dose-resIxmse relationship in numerous cases; and (3) the biologic plausibility given the increased risk of respiratory symptoms in adult smokers. (The Health Assessment at 5-18). The Agency's reasoning is misguided. There is no reason to require the demonstration of a single source of bias or confounding to explain an observed association before concluding that it may be artifactual. There are many potential sources of bias and confounding in the studies that may act independently to produce the weak and inconsistent observed associations. The odds ratios reported in the studies are generally under 2.0 and quite variable. The particular symptoms or illnesses associated with parental smoking vary from study to study and no single symptom or illness is found in over half of the studies. Some studies do show an increase in symptom prevalence with an increasing number of smokers in the home but the effects of bias and confounding could also increase with the number of smokers and produce the appearance of a dose-response. The Health Assessment acknowledges the possibility of _re¢_~l bias in the studies but states that it is unlikely "to have a broad-based influence across numerous studies." (1'he Health Assessment at 5-13). Again, no evidence supporting this assertion is described. EPA claims that the evidence is substantial and that if ETS exposure had no effect on respiratory symptoms, then an increase or decrease of prevalence with ETS exposure would be equally likely. (The Health Assessment at 5-13). This assertion ignores bias, confounding and chance as possible explanations. Moreover, publication bias and statistical multiple comparisons may explain the slight predominance of observed elevated odds ratios. In essence, EPA accepts uncritically the epidemiologic observations and makes no attempt to 71

evaluate the impact of these factors. The Agency has not met its burden to conduct a reasoned analysis and present the scientific basis and rationale for its conclusion. (51 Fed. ~ 33992). (ii) Acute Lower Resoiratorv Tract Illness. The Health Assessment concludes that parental smoking is associated with acute respiratory tract illness (bronchitis, pneumonia) because the observations across the cumulative studies cannot be statistically attributed to chance alone. (The Health Assessment at 5-24). No statistical evidence is offered in support of this assertion. The. Agency concedes that potential sources of bias and confounding have not been adequately assessed to explain the nature of the association. The NRC, 1986, concluded that the observed associations may be indirect (non-causal) and attributable to confounding influences. Conh"ol for parental respiratory status reduces significantly the observed associations. (Colley et a/. 1974; Leeder et al. 1976a,b; Schenker et a~ 1983; Ware et al. 198~). The EPA is correct that the observed associations are unexplained and that the data are insufficient to reach a conclusion regarding the influence of early respiratory illness on respiratory health in later life. (The Health Assessment at 5-24 to 5-25). The Agency's overall narrative on respiratory illness is so one-sided and cursory, however, that it cannot be said to have any impact on informing the public about the issue. (iii) Pulmonary_ Function. The Health Assessment concludes that parental smoking is associated with decreased lung function in childhood and with a small reduction in the rate of pulmonary growth and development. (The Health Assessment at 5-32). As the Agency suggests, the observed associations have been inconsistent with respect to sex and parameter affected. The Agency notes the difficulty in measuring lung function in small children and of

attempting to isolate any potential effect of ETS exposure from the major determinants of lung function and concedes that the overall evidence is difficult to assess. (The Health Assessment at 5-31). EPA's review does nothing to bring any clarity to the issue. The NRC also d~bed the difficulty of measuring lung function in small children and of attempting to isolate any potential association with parental smoking from the major determinants of lung function. It concluded that the overall evidence is difficult to assess and that it is not possible to determine whether the observed associations are direct or indirect. Finally, the NRC notes that the small decrements in FEVt_are unlikely to be clinically significant. The possibility of chance variations, genetic factors, and the influence of motivation, learning and socioeconomic status is controlled poorly in these studies. (Witorsch and Witorsch, 1989). Spirometric performance must be standardized for a child's age, sex and height. (Bates, 1989). Performance in forced vital capacity tests may be affected by the technician directing the tests, the type of spirometer used, and the subject's effort. (Bates, 1989). Those factors must be assessed for each study. The Agency's conclusion that ETS exposure is associated with decreased lung function in childhood is purely arbitrary. In the two pages of analysis devoted to the topic, EPA finds that: (1) lung function is difficult to measure in children; (2) factors of familial resemblance complicate the analysis; (3) any effect of ETS exposure is likely to be small; (4) two major longitudinal studies found conflicting results; (5) differences between exposure levels and host-related variables make it difficult to assess the overall data, and; (6) it is not possible to determine whether ETS is causing the decreased lung function or whether increased infection 73

rates in the children of smokers are responsible for the observed decreases. With no explanation of how these deficiencies are outweighed by countervailing data or considerations, the Agency concludes that cumulative evidence demonstrates an association. No reasoned weighing of the data using established criteria is described. (The Health Assessment at 5-31 to 5-32). Without more, the Agency's conclusion is afoitrary and will mislead the public. (iv) ~~,~[~. EPA reviews respiratory disorders other than respiratory symptoms, acute ili.nesses and pulmonary function changes_under the heading of "related results." The Agency correctly determines that no association between maternal smoking and childhood asthma has been shown. (The Health Assessment at 5-33). Less than half of the published studies report statistically significant associations. The majority report no elevated associations. The Health Assessment concludes that parental smoking is associated with an increased incidence of middle ear effusion in children. HPA notes; however, that no mechanism is known and that confounding may explain the association. The Agency has fallen far short of providing a reasoned analysis supporting its conclusion. Again, its conclusion is arbitrary and without foundation. B. The A_~enc_v's Review Will Mislead The Public EPA's Risk Assessment Guidelines require that Agency scientists "identify the strengths and weaknesses of each assessment by describing uncertainties, assumptions and limitations, as well as the scientific basis and rationale for each assessment." (51 l~ed. Reg. 33992). The Agency's review of parental smoking and childhood respiratory health fzilg to

do so. The Health Assessment includes irrelevant and misleading material. The epidemiologic data are overinterpreted and flaws are minimized. The studies are not critically reviewed and confounding variables are discounted as being unable to explain completely the observed associations. Association is not distinguished clearly from causation. Instead, the Health Assessment blurs the distinction in a fashion impenetrable by the lay public. The Agency should state expressly that no causal conclusions can be made and that no regulation of smoking - based on protecting children's health - is justified scientifically. (i) Mainstream smokin~ and COpD. EPA suggests that a growing knowledge base implicating smoking as a cause of chronic obstructive pulmonary disease ('COPD') increases the likelihood that ETS exposure is related to respiratory disorders. (The Health Assessment at 5-I). The Health Assessment contains no review of the data and only one reference to relevant literature is made. No scientific analysis is presented. Smoking has not been established as a cause of COPD. No animal model has confirmed the epidemiologic obse~ations. Moreover, smoking data are irrelevant to ETS and health: the physio-chemical differences between MS and ETS coupled with the physiological and quantitative exposure differences prevent comparison. Discussion of smoking data should be deleted. (ii) ETS and adult re~iratory_ health. The Agency restricts its analysis of ETS and respiratory disorders to children. However, contrary to the Agency's assertion at 5-I, several studies in adults have been reported. In a 1989 review, Witorsch identified fifteen epidemiologic studies of ETS and pulmonary function and/or respiratory symptoms or illness in adults and five studies of acute ETS exposure in asthmatic adults. (Witorsch, 1989). The studies produced inconsistent results and no proof that ETS exposure is associated with 75

impaired respiratory health or pulmonary function in nonsmoking adults. The acute exposure studies generally failed to produce an adverse change in pulmonary function even under extreme exposure conditions. These studies undercut the plausibility of obser~g an important effect in children. The Agency's attempt to minimize the studies is further evidence of a biased approach. ~" EPA suggests that a strength of the children's studies is that children are less exposed socially to potential confounding agents. (The Health Assessment at 5-I). This statement is overly simplistic. Children's lifestyles and exposures are dictated by those of their parents. The effects of parental socioeconomic class on children's respiratory health was examined by Kerigan et a/., 1986. They found parental smoking to be inversely related to family income and positively correlated with poorer outdoor air quality, increased parental coughing, higher gas stove usage, frequent change of address and lower per capita living space. Socioeconomic status includes several components and its adjustment in epidemiologic studies is complex. (Witorsch, 1990). Restricting a study to infants is no guarantee that socially encountered confounders will be eliminated. The Agency has given inadequate consideration to socioeconomic status and related variables as confounding factors. Factors such as outdoor air quality, home heating, air conditioning, and humidity must be examined. CvVitorsch, 1990). Studies of parental smoking and childhood respiratory health adjust infrequently for attendance at day care centers. Children who attend day care centers have been reported to incur an increased incidence of respiratory disorders. (Strangert, 1976; Stahlberg, 1980; Haskins and Kotch, 1986; Fleming et oL, 1987). If parental smoking is correlated with use

of day care fa~lities, spurious association for ETS might be produ~xi. In addition, family size and the presence of older siblings have been reported to increase the incidence of respiratory symptoms and disease in young children. (Colley, 1971; McCormochie and Roghrnann, 1986). (iv) ~]~[y_~]g~. The Health Assessment fails to convey to the public the overarching, methodologic flaws in the epidemiologic studies. All employed essentially a similar design. Family health information and parental smoking history are obtained using a standardized questionnaire. The response usually provides-~he sole source of information. If pulmonary function was examined, spirometric measurements were made. The incidence of respiratory disorders or pulmonary function was then compared statistically according to parental smoking stares. This design is valid only if parental smoking is an unbiased, nonconfounded surrogate of ETS exposure. It is not. The Agency must quantify the con~bution of bias and confounding before it can conclude that ETS is associated with respiratory disorders. The Agency's discussion of childhood exposure is unbalanced. Nicotine is not ETS. Contrary to the EPA's assertion at 5-4, cotinine studies do not demonstrate absorption, distribution and metabolism of ETS by infants. They demonstrate only that nicotine from ETS or the diet has been absorbed. No study of ETS and children's health has employed an exposure biomarker that is potentially relevant to pathogenesis. There is no evidence that cotinine is the etiologic agent or linearly related to any risk-relevant compounds for middle ear effusion or any other disorder. As discussed in Section I of these comments, cotinine concentrations provide little basis for assessing ETS exposure. 77

C. Scientific and Political Issues Should Be S _e_e~rated Clearly The review is designed to persuade the lay public to adopt the Agency's antismoking agenda - not to objectively inform the public about the scientific evidence. The Health Assessment purports to be a scientific review, but is in fact driven by the Agency's policy preferences. The Agency's scientific analysis is subjugated to its antismoking bias: the Health Assessment concludes that no causal association between ETS and childhood respiratory disorders has been established, but that it is prudent and reasonable to treat ETS as though one had. (The Health Assessment at 1-9). In fact, the Agency's analysis does not demonstrate that the studies show an association (much less causal association) between ETS and these disorders that is unbiased and unconfounded. smoking for policy reasons, it should do so expressly. If the Agency wishes to discourage EPA should not, however, exaggerate claims about ETS and children's health to bias the public against smoking. 78

IV. ~ EPA has misused the process of risk analysis by preparing the Health Assessment for the apparent purpose of furthering an antismoking agenda - not for scientifically evaluating Ers and health. The Health Assessment violates the Agency's own requirement that scientific analysis be clearly separated from policy formation. As written, the Health Assessment is analogous to a lawyer's brief. The strongest possible case is made to discourage smoking and to encourage the tightening of smoking restrictions. Data regarding ETS and health are. overinterpreted and presented in a bi _as~d fashion. Crucial assumptions are unstated and substantial uncertainty is not disclosed. Conflicting data or analyses are ignored. No examination of the individual studies for quality is undertaken. As a result, poor quality data supporting EPA's conclusions are given weight while data detracting from the Agency's findings are minimized or completely ignored. Myriad weaknesses and limitations of the Health Assessment are not discussed. The Health Assessment deviates from the Agency's own guidelines and bases its conclusions on arbitrary statistical analyses with no substantial evidentiary basis. In short, the Health Assessment is EPA's attempt to develop a scientific rationale to support a predetermined conclusion to advocate smoking restrictions. The HIS and lung cancer analysis is little more than a statistical manipulation of a selected subset of epidemiologi¢ studies of spousal smoking and lung cancer. The Health Assessment contains no discussion of the physico-chemical properties of ETS and no critical examination of the relevance of mainstream smoke data -- despite the fact that the assumption that mainstream and ETS are equivalent is critical to the Agency's b:~'~rd identification. It contains no summary of metabolic or pharmacokinefi¢ data. Data from animal inhalation

studies - information required to draw a causal inference - are ignored by the Agency. The lack of any mechanistic understanding of the observed statistical association is not addressed. The epidemiologic studies of spousal smoking and lung cancer are not examined critically. The role of confounding is dismissed with no analysis. EPA uses a statistical procedure to combine the studies in an effort to overcome their lack of individual statistical power. This combination is inappropriate: (1) the basic principle that studies must be comparable in design, conduct and analysis to be combined is violated in several ways, (2) the studies combined are not representative of all studies of the topic, and O) studies of poor quality were included because no clear set of criteria for including and excluding studies was applied. The Health Assessment advances from a statistical analysis to the conclusion that ETS is a Group A carcinogen based on entirely conclusory statements. No discussion of criteria for drawing causal inferences is presented. In fact, the epidemiologic data are weak and over 80% of the individual studies are statistically insignificant. It is clear that major sources of confounding have not been controlled for. Multiple bias sources are not corrected in data analysis. The Agency's attempt to pass off a test for trend as a dose-response analysis is unconvincing. Under no analysis is a statistically significant dose-response relationship observed. The epidemiologic data are, on the whole, inconsistent and unconfirmed by animal data. No basis exists for extrapolating the data on nonsmoking wives to men and ex- smokers. The Agency has not demonstrated that ETS causes lung cancer. Instead, it has constructed a "straw man" rationale to validate its advocacy of smoking restrictions. Because the Agency has not demonstrated that ETS exposure causes lung cancer, the mortality 80

projections are baseless and should be deleted. The Agency violates overtly its obligation to separate risk assessment and risk management issues in its review of ETS and non-malignant respiratory disease in children. The Agency concedes that the epidemiologic data are weak, confounded and do not prove that ETS causes respiratory disease in children. Yet, EPA concludes that it is reasonable and prudent to treat ETS as though the proof were adequate. This blatant politicizafion of ETS and children's health does a disservice to the public interest. Exaggeration of scientific data to support policy preferences is not prudent and reasonable - it is intellectually dishonest and an abuse of EPA's discretion. The scientific data should be presented accurately to the public. Policy preferences should not be foisted upon the lay public disguised as science. If the agency wants to present its political agenda to the public, let it do so squarely and let the public decide the issue with scientific analysis clearly separated from policy preferences.

BIBLIOGRAPHY Adami, H.-O., Persson, I., Hoove~, R., Schairer, C., and Bergkvist, L., "Risk Of Cancer In Women Receiving Hormone Replacement Therapy," Int J Cancer~ 44:833-839, 1989. Adlkofer, F., Scherer, G. Wenzel-Hartung, R., Brune, H. and Thomas, C., "Exposure of Hamsters and Rats to Sidestream Smoke of Cigarettes: Preliminary Results of a 90-Day- Inhalation Study," in R. Perry and P. W. Kirk (eds.), Indoor and Ambient Air Ouali _ty, Selper, Ltd., London, 1988, pp. 252-258. Ahlborn, W., Uberla, K., "Passive Smoking and Lung Cancer:. Reanalyses of Hirayama's Data," in R. Perry and P. W. Kirk (eds.), Indoor and Ambient Air Ouality, Selper, Ltd., London, 1988, pp. 169-178. Akiba, S., Kato, H:, Blot, W. J., "Passive Smoking and Eung Cancer Among Japanese Women," Can Res 46:4804-4807, 1986. Alavanja, M. C. R., Brownson, R., Boice, J. D., "Risk Factors for Lung Cancer Among Nonsmoking Women," Socie_ty for Epidemiologic Research, Snowbird, Utah, 1990. Albanes, D., Blair, A. and Taylor, P. R., "Physical Activity and Risk of Cancer in the NHANES I Population," AJPH 79:744-750, 1989. Axelson, O., Andersson, K., Desai, G., Fagedund, I., Jansson, B., Karlsson, C., Wingren, G., "Indoor Radon Exposure and Active and Passive Smoking in Relation to the Occurrence of Lung Cancer," Scand J Environ H~lth, 14:286-292, 1988. Bates, D. V., "Respiratory Function in Disease," 3rd ed., W. B. Saunders, Philadelphia, 1989. Begg, C. B., Berlin, J. A., "Publication Bias and Dissemination of Clinical Research," JNCI, 81:107-115, 1989. Biber, A., Scherer, G., Hoepfner, I., Adlkofer, F., Heller, W.D., Haddow, J.E. and Knight, G., "Determination of Nicotine and Cotinine in Human Serum and Urine: An Interlaboratory Study," ~f~qg.L[,g~, 35:45-52, 1987. Blot, W. J. and Fraumeni, J. F. "Passive Smoking and Lung Cancer," JNCI, 77:993-1000, 1986. Brown, S. M., Selvin, S., and Winkelstein, Jr., W., "The Association Of Economic Status With The Occurrence Of Lung Cancer," Cancer, 36:1903-191 I, 1975. Brownson, R. C., Reif, J. S., Keefe, T. J., Ferguson, S. W., Pfitzl, J. A., "Risk Factors for 82

Adenocarcinoma of the Lung," American Journal of Epidemiology, 125:25-34, 1987. Buffler, P. A., Pickle, L. W., Mason, T. J., Contant, C., "The Causes of Lung Cancer in Texas," in M. Mizell and P. Correa (eds.), Lung Cancer: Causation and Prevention, Verlag Chemie International, Louisiana, 1984, pp. 83-99. Butler, T. L., "The Relationship of Passive Smoking to Various Health Outcomes Among Seventh Day Adventists in California:" A Disserlation Submitted in Partial Satisfaction of the Requirements for the Degree, University of California, 1988. Castro, A., Monji, N., "Dietary Nicotine and Its Significance in Studies on Tobacco Smoking," Biochemical Archives, 2:91-97, 1986. Chalmers, T. C., Frank, C. S., Reitman, D., "Minimizing the Three Stages of Publication Bias," JAMA, 263:.1392-1395, 1990. Chart, W. C., Colbourne, M. J., Fung, S. C., Ho, H. C., "Bronchial Cancer in Hong Kong 1976-1977," British Journal of Cancer, 39:182-192, 1979. Colley, J. R. T., "Respiratory Disease In Childhood," Br Med Bull, 27:9-14, 1971. Colley, J. R. T., Holland, W. W., Corkhill, R. T., "Influence of Passive Smoldng and Parental Phlegm on Pneumonia and Bronchitis in Early Childhood," Lancet, 2:1031-1034, 1974. Colley, J. R. T., "Respiratory Symptoms in Children and Parental Smoking and Phlegm Production," Br Med J, 2:201-204, 1974. Correa, P., Pickle, L. W., Fontham, E., Lin, Y., Haenszel, W., "Passive Smoking and Lung Cancer," Lariat, 2:595-597, 1983. Coultas, D. B., Howard, C. A., Peake, G. T., Skipper, B. J. and Samet, J. M., "Salivary Cotinine Levels and Involuntary Tobacco Smoke Exposure in Children and Adults in New Mexico," Am Rev Respir Dis, 136:305-309, 1987. Cummings, M. K., Markello, S. J., Mahoney, M., Bhargava, A. K., McElroy, P. D., Marshall, J. R., "Measurement of Current Exposure to Environmental Tobacco Smoke," Archives of Environmental Health, 45:74-79, 1990. Curvall, M., Vala, E. K., Englund, G., and Enzell, C. R., "Urinary Excretion of Nicotine and its Major Metabolites," 43rd Annual Tobacco Chemists' Research Conference, Richmond, VA, 1989. Dalager, N. A., Pickle, L. W., Mason, T. J., Corre~, P., Fontham, E., Stemhagen, A., 83

Buffler, P. A., Ziegler, R. G., Fraumeni, ~. F., "The Relation of Passive Smoking to Lung Cancer," Cancer Research, 46:4808-4811, 1986. deBethizy, J. D., Bates, L. E., Davis, R. A., Heavner, D. L., Nelson, P. R., Walker, J. C., and Robinson, J. H., "Nicotine Absorption in Humans Following Exposure to Environmental Tobacco Smoking Generated From Different Types of Cigarettes," In Present and Future of Indoor Air Ouality, Excerpta Medica Int. Cong. Set. 860:269-276, 1989. Denton, F. T., "The Effects of Publication Selection on Test Probabilities and Estimator Distributions," Risk Analysis, 10:131-136, 1990. Dickersin, K., "The Existence of Publication Bias and Risk Factors for Its Occurrence," JAMA, 263:1385-1389, 1990. Dickersin, K., Chag., S., Chalmers, T. C., Sacks, H.S., S.mith, H., "Publication Bias and Clinical Trial," Controlled Clin Trials, 8:343-353, 1987. Doll, R., "The Aetiology of the Spanish Toxic Shock Syndrome: Interpretation of the Epidemiological Evidence," Report to the WHO Regional Office for Europe, 1985. Edling, C., Kling, H. and Axelson, O., "Radon In Homes - A Possible Cause Of Lung Cancer," Scand J Work Environ Health, I0:25-34, 1984. Etzel, R. A., "A Review of the Use of Saliva Cotinine as a Marker of Tobacco Smoke Exposure," Preventive Medicine, 19:190-197, 1990. Eudy, L. W., Thome, F. A., Heavner, D. L., Green, C. R. and Ingebrethsen, B. J., "Studies on the Vapor-Particulate Phase Distribution of Environmental Nicotine by Selective Trappings and Detection Methods," Proc. of the Air Pollution Control Association 79th Annual Meeting, 1986, pp. 2-14. Fleming, D. W., Cochi, S. L., Hightower, A. W., Broome, C. V., "Childhood Upper Respiratory Tract Infections: To What Degree is Incidence Affected by Day-Care Attendance7," Pediatrics, 79:55-60, 1987. Friedman, G. D., Petitti, D. B., Bawol, R. D., "Prevalence and Correlates of Passive Smoking," Am J Public Health, 73:401-405, 1983. Gao, Y., Blot, W. J., 7_.heng, W., Ershow, A. G., Hsu, C. W., Levin, L. U., Zhang, R., Faumeni, J. F., "Lung Cancer Among Chinese Women," International Journal of Cancer, 40:604-609, 1987. Garfinkel, L., "Time Trends in Lung Cancer Mortality Among Nonsmokers and a Note on Passive Smoking," JNCI, 66:1061-1066, 1981.

Garfinkel, L., Auerbach, O., Joubert, L., "Involuntary Smoking and Lung Cancer: A Case- Control Study," JNCI, 75:463-469, 1985. Geng, G-Y., Liang, Z. H., Zhang, A. Y., Wu, G. L., "On the Relationship Between Smoking and Female Lung Cancer," Smokin~ and Health 1987, Excerpta Medica, New York, 1988, pp. 483-486. Glass, G. V. "Primary, Secondary, and Meta-Analysis of Research," Educational Researfb, 5:3-8, 1976. Glass, G. V., McGaw, B., Smith, M. L., "Meta-Analysis in Social Research," Beverly Hills, CA: Sage Publishing, 1981. Goodman, Marc T., Kolonel, L. N., Yoshizawa, C. N., and Hankin, J. H., "The Effect of Dietary Cholestorot and Fat on the Risk of Lung Cancer in Hawaii," Am J Epidemiology, 125:1241-1255, 1988. Gori, G. B., "A Summary Appraisal," in G. B. Gori and F. G. Book (eds.), Banbury R _e~on 3: A Safe Cigarette, Cold Spring Harbor Laboratory, New York, 1980, p. 355. Grundmann, E., Muller, K-M., Winter, K. D., Letter to the Editor, British Medical Journal, 282:1156, 1981. Haley, N., "Sidestream Smoke Uptake By Syrian Golden Hamsters in An Inhaltion Bioassay," Indoor Air '87: Proceedings of An International Conferenf.f ¢?n Indoor Air Ouality and Climate, 2:68-73, 1987a. Haley, N. J., Adams, J. D., Alzofon, J. and Hoffmann, D., "Uptake of Sidestream Smoke By Syrian Golden Hamsters," Toxicology Letters, 35:83-88, 1987"o. Harris, J. E., DuMouchel, W. H., Letter to the Editor, British Medical Joumal, 283:915, 1981. Haskins, R., Koteh, J., "Day Care and Illness: Evidence, Costs, and Public Policy," Pediatrics, 77(Suppl):951-982, 1986. Hayes, R. B., "Lung Cancer in Motor Exhaust-Related Occupations," American Journal of Industrial Medicine, 16:685-695, 1989. He, Z., Chen, W., Liu, Z., Chapman, R.S., "An Epidemiologieal Study of Lung Cancer in Xuan Wei County, China: Current Progress (I)," Second Annual Meeting of the International Society for Environmental Epidemiology, Berkeley, CA, Aug. 13-15, Abstract, 1990. 85

Hedges, L. V., Olkin, L., "Statistical Methods for Meta-Analysis,~ N~vv York: Academic Press, 1985, pp. 1-361. Heller, W. D., "Hirayama's Data Are Not Representative," Munch Med Wschr, 124:11-12, 1982. Hill, A. B., "A Short Textbook of Medical Statistics," Hodder and Stoughton, London, 1977, pp. 281-297. Hiller, F. C., Mazumder, M. K., Wilson, J. D., McLeod, P. C. and Bone, R. C., "Human Respiratory Tract Deposition Using Multinodal Aerosols," J Aerosol Sfi, 13:337-343, 1982. I-Iirayama, T., "Lung Cancer in Japan: Effects of Nutrition and Passive Smoking," International Lung, Cancer U_t~date Conf~l~rlfe, New Orleans, LA, March 3-5, 1983. Hoffmann, D., Brunnemann, K., Adams, J., Rivenson, A., Hecht, S., "N-Nitrosamines in Tobacco Carcinogenesis," Banbury R _eport 2, 1982 at 224. Hole, D. J., Gillis, C. R., Chopra, C., Hawthorne, V. M., "Passive Smoking and Cardiorespiratory Health in a General Population in the West of Scotland," ~ Journal, 299:423-427, 1989. Humble, C. G., Samet, J. M., Pathak, D. R., "Marriage to a Smoker and Lung Cancer Risk," Am J Publ Health, 77:598-602, 1987. Hunter, J. E., Schmidt, F. L., Jackson, G. B., "Meta-Analysis: Cumulating Research Findings Across Studies," Beverly Hills, CA: Sage Publishing, 1982, pp. 1-176. Hunter, J. E., Schmidt, F. L., "Methods of Meta-Analysis: Correcting Bias in Research Findings," Newbury Park: Sage Publishing, 1990, p. 506. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Tobacco Smoking, Volume 38, Lyon, 1986. Idle, J. R., "Titrating Exposure to Tobacco Smoke Using Cotinine - A Minefield of Misunderstanding," Journal of Clinical Epidemiology, 43:313-317, 1990. Ingebrethsen, B. J., "The Physical Properties of Mainstream Cigarette Smoke and Their Relationship to Deposition in the Respiratory Tract," Extrapolation of Dosimetric Relationships For Inhaled Parti¢l¢~, Academic Press, Inc., Chapt. 12:125-141 and references cited therein, 1989. Inoue, R. and Hirayama, T., "Passive Smoking and Lung Cancer in Women," In: ~ ~, Elsevier Science Publishers, pp. 283-285, 1988. 86

Ives, J. C., Buffler, P. A., Selwyn, B. J., Hardy, R. J., Decker, M., "Lung Cancer Mortality Among Women Employed in High-Risk Industries and Occupations in Harris County, Texas, 1977-1980," American Journal of E_oidemiol _ogy_, 127:6:5-74, 1988. Jain, M., Burch, J. D., Howe, G. R., Risch, H. A., and Miller, A. B., "Dietary Factors and Risk of Lung Cancer:. Results from a Case-Control Study, Toronto, 1981-1985," Int J C~cer, 45:287-293, 1990. Jarvis, M. J., Tunstall-Pedoe, H., Feyerabend, C., Vesey, C. and Salloojee, Y., "Biochemical Markers of Smoke Absorption and Self-Reported Exposure to Passive Smoking," J Evidemiol Commun Health, 38:335-339, 1984. Jarvis, M. J., McNeill, A. D., Russell, M. A. H., West, R. J., Bryant, A. and Feyerabend, C., "Passive Smoking in Adolescents: One-Year Stability of Exposure on the Home," Lariat, i: 1324-1325, 1987. -. Jarvis, M. L., "Application of Biochemical Intake Markers to Passive Smoking Measurement and Risk Estimation," Mutation Research, 222:101-110, 1989. Kabat, G. C., Wynder, E. L., "Lung Cancer in Nonsmokers," Cancer, 53:1214-1221, 1984. Katada, H., Mikami, R., Konishi, M., Koyama, Y., Narita, N., "Effect of Passive Smoking in Lung Cancer Development in Women in the Nara Region," Gan No Rin~ho, 34:21-27, 1988. Kerigan, A. T., Goldsmith, C. H., Pengelly, L. D., "A Three-Year Cohort Study of the Role of Environmental Factors in the Respiratory Health of Children in Hamilton, Ontario," Am Rev Resp_ ir Dis, 133:987-993, 1986. Kilpatrick, S. J., "An Example of Extra-Poisson Variation Suggesting an Under-Specified Model," in C. J. Bieva, Y. Courtois and M. Govaerts (eds.), Present and Future of Indoor ~r...Qllal~, Excerpta Medica, Amsterdam, 1989, pp. 83-90. Kilpatrick, S. J., Viren, J., "Age as a Modifying Factor in the Association Between Lung Cancer in Non-Smoking Women and Their Husbands Smoking Status," in R. Perry and P.W. Kirk (eds.), Indoor and Ambient Air Ouality, Selper, Ltd., London, 1988, pp. 195-202. KOO, L. C., Ho, J. H-C., and Saw, D., "Is Passive Smoking And Added Risk Factor For Lung Cancer In Chinese Women'?.," J Exp Clin Cancer Re~, 3:277-283, 1984. Koo, L. C., "Environmental Tobacco Smoke and Lung Cancer: Is It The Smoke Or The Diet.'?," in C. J. Bieva, Y. Courtois and M. Govaerts (eds.), Present and Future of Indoor Air Ouali _ty, Excerpta Medica, Amsterdam, pp. 65-75, 1989. 87

Koo, L. C., Ho, J. H=C., Saw, D., Ho, C=Y., "Measurements of Passive Smoking and Estimates of Lung Cancer Risk Among Nonsmoking Chinese Females," Int J Cancer, 39:162- 169, 1987. Koo, L. C., Ho, L H-C., Saw, D., =Active and Passive Smoking Among Female Lung Cancer Patients and Controls in Hong Kong," J Exp Clin Cancer Res, 4:367-375, 1983. Koo, L. C., Ho, J. H-C., Lce, N., "An Analysis of Some Risk Factors for Lung Cancer in Hong Kong," Int ~ Caner, 35:149-155, ,1985. Komegay, H. R., Kastenbaum, M. A., Letter to the Editor, British Medical Jourrla], 283:914, 1981. Kulessa, C. H. E., Blohmke, M., Jagschitz, P., Stelzer, O., Cooper, C. L., and Eysenck, H. J., "Pshycosocial Personality Traits and Cigarette Smoking Among Bronchial Carcinoma Patients," Stress Medicine, 5:37-46, 1989: Kvale, G., Bielke, E., and Heuch, I., "Occupational Exposure and Lung Cancer Risk," Int I Cancer, 37:185=193, 1986. Lam, T. H., Kung, J. T. M., Wong, C. M., Lam, W. K., Klcevens, ~. W. L., Saw, D., Hsu, C., Senevirame, S., Lam, S. Y., Lo, K. K., Chan, W. C., "Smoking, Passive Smoking and Histological Types in Lung Cancer in Hong Kong Chinese Women," Brit J Cancer, 56:763-678, 1987. Lange, P., Nyboe, J., Appleyard, M., Jensen, G., and Schnohr, P., "Ventilatory Function and Chronic Mucus Hypersecretion as Predictors of Death from Lung Cancer," Am R~ Respir Dis 141:613-617, 1990. Layard, M. W., Viren, J. R., "Assessing the Validity of a Japanese Cohort Study," in C. J. Bieva, Y. Courtois and M. Govaerts (eds.), Present and Future of Indoor Air Ot~ali _ty, Excerpta Medica, Amsterdam, 1989, pp. 177-180. Le Marchand, L., Yoshizawa, C., Kolonel, N., Hankin, J., and Goodman, M., "Vegetable Consumption and Lung Cancer Risk: A Population-Based Case-Control Study in Hawaii," JNCI, 81:1158-1164, 1989. Leaderer, B. P., "Assessing Exposures to Environmental Tobacco Smoke," Risk Analysis, 10:19-26, 1990. Lee, P. N., Chamberlain, J., Alderson, M. R., "Relationship of Passive smoking to Risk of Lung Cancer and Other Smoking-Associated Diseases," British Medical Journal, 54:97-105, 1986. 88

Lee, P. N., "An Alternative Explanation for the Increased Risk of Lung Cancer in Non- Smokers Married to Smokers," in R. Perry and P.W. Kirk (eds.), Indoor and Ambient Air Ouality, Selper Ltd., London, 1988, pp. 149-158. Lee, P. N., Letter to the Editor, British Medical Jour0al, 283:1465-1466, 1981. Lee, P. N., "Passive Smoking and Lung Cancer Association: A Result of Bias7," Human Toxicol, 6:517-524, 1987. Leeder, S. R., Corkhill, R. T., Irwig, L. M., Holland, W. W., Colley, I. R. T., "Influence of family factors on the Incidence of Lower Respiratory Illness During the First Year of Life," Br J Prey Soc Med, 30:203-212, 1976a. Leeder, $. R., Corkhill, R. T., Wysocki, M. L, Holland, W. W., Colley, J. R. T., "Influence of Personal and Family Factors on Ventilatory Function of Children," Br J Prey Soc Meal 30:219-224, 1976b. Lees, R. E. M., Steele, R., and Roberts, J. H., "A Case-Control Study Of Lung Cancer Relative To Domestic Radon Exposure," Int J Epidemiology, 16:7-12, 1987. Letzel, H., "Passive Smoking and Lung Cancer: Methods for the Demonstration of Causality and for the Determination and Validation of Exposure," in K. Ubefla, O. Rienhoff and N. Victor (eds.), Medical Information Science and Statistics, 1988. Lloyd, O. L. I., Ireland, E., Tyrrell, H., Williams, F., "Respiratory Cancer in a Scottish Industrial Community: A Retrospective Case-Control Study," J Soc Occup M¢~t, 36:2-8, 1986. MacDonald, E. J., Letter to the Editor, British Medical Journal, 283:1465, 1981. Machacek, D., and :Iiang, N., "Quantification of Cotinine in Plasma and Saliva by Liquid Chromatography," Clin Chem, 32:979-982, 1986. Mann, C., "Meta-Analysis in the Breech," Science, 249:476-481, 1990. Mantel, N., Letter to the Editor, British Medical Journal, 283:914-915, 1981. Mantel, N., "Epidemiologic Investigations: Care in Conduct, Care in Analysis, and Care in Reporting," Cancer Res Clin On¢ol, 105:113-116, 1983. Matsunga, S. K., Pleizia, P. M., Karol, M. D., Katz, M. D., Camilli, A. E. and Benowitz, N. L., "Effects of Passive Smoking on Thcophylline Clearance," Clin Pharmacol Thor, 46:399-407, 1989. 89

McConnochie, K. M., Roghmann, K. J., "Parental Smoking, Presence of Older Siblings, and Family History of Asthma Increase Risk of Bronchiolitis," Am J Dis Children, 140:806-812, 1986. Mettlin, C. J., Schoenfeld, E. R., and Natarajan; N., "Patterns of Milk Consumption and Risk of Cancer," Nu~ Can~, 13:89-99, 1990. Mumford, J. L., He, X. Z., Chapman, R. S., Cao, S. R., Harris, D. B., Li, X. M., Xian, Y. L., Jiang, W. Z., Xu, C. W., Chuallg, J. C., Wil~on, W. E. and Cooke, M., "Lung Cancer and Indoor Air Pollution in Xuan Wei, China," Science, 235:217-220, 1987. Natioml Research Council (NRC) (1983) Risk Assessment in the Federal Government: Managing the Process, National Academy Press, Washington, D.C. National Research Council (NRC) (1986) Environmental Tgbacco Smoke: Measuring Exposures and Assessing Health Effects, National Academy Press, Washington, D.C. National Research Council OqRC) (1985) Committee on the Epidemiology of Air Pollution. Epidemiology and Air Pollution, National Academy Press, Washington, D.C. National Research Council (NRC) (1988) Complex Mixtures: Methods for In Vivo Toxicity Testing, National Academy Press, Washington, D.C. Nelson, P. R., Heavner, D. L. and Collie, B. B., "Characterzafion of the Environmental Tobacco Smoke Generated by Different Cigarettes," In Present ~nd Futur~ of Indoor Air Ouality, Excerpta Medica Int. Cong. Ser. 860:277-282, 1989. Nelson, P. R., Ogden, M. W., Maiolo, K. C., Heavner, D. L. and Collie, B. B., "Predictive Value of Nicotine as an Environmental Tobacco Smoker Marker," in Proceedings of "Indoor Air '90: The Fifth International Conference on Indoor Air Quality and Climate," Inglewood Printing Plus, Aurora, ON, 2:267-272, 1990b. Nelson, P. R., Heavner, D. L., and Oldaker III, G. B., "Problems With the Use of Nicotine as a Predictive Environmental Tobacco Smoke Marker," Proceedings of the 1990 EPA/A&WMA International Symposium: Measurement of Toxic and Related Air Pollutants, Air & Waste Management Association, 1990, Pittsburgh, to appear 1990a. Office of Science & Technology Policy, "Chemical Carcinogenesis: A Review of the Science and its Associated Principles," ~.¢~,J:i~.._~,i~[~, March 14, 1985. Ooi, W. L., Elston, R. C., Chert, V. W., Bailey-Wilson, J. E., and Rothschild, H., "Increased Familial Risk for Lung Cancer," JNCI, 76:217-222, 1986. Pershagen, G., Hrubec, Z., Svensson, C., "Passive Smoking and Lung Cancer in Swedish

Women," ~d~, 125:17-24, 1987. Perusse, L., I.e.blanc, C., Tremblay, A., Allard, C., Theriault, G., Landry, F., Talbot, J. and Bouchard, C., "Familial Aggregation in Physical Fitness, Coronary Heart Disease Risk Factors, and Pulmonary Function Measurements," Prey Meal, 16:607-615, 1987. Pollack, E. S., Nomura, A. M. Y., Heilbrun, L. K., Stemmermann, G. N. and Green, S. B., "Prospective Study of Alcohol Consumption and Cancer," ]~,t~g_J.~, 310:617-621, 1984. Proctor, C. J., Wal'refl, N. D., Bevan, M. A. J., and Baker-Rogers, J., "A Comparison of Methods of Assessing Exposure to Environmental Tobacco Smoke in Non-Smoking British Women," Submitted to Environmental International, 1990. Reynolds, P., Kaplan, G. A., Cohen, R. D., "Passive Smoking and Cancer Incidence: Prospective Evidence From the Alameda County Study," Presented at the Annual Meeting of the Society for Epidemiologic Research, Amherst, MA, June 16-19, 1987. Rickert, W. S., "Some Considerations When Estimating Exposure to Environmental Tobacco Smoke (ETS) with Particular Reference to the Home Environment," Can J Pub Health, 79:$33-$39, 1988. Rutsch, M., "Inadequate Statistical Substantiation," from Munch Med Wschr, 124:1-3, 1982. Rutsch, M., Letter to the Editor, British Medical Journal, 282:985, 1981. Rylander, R., Haussmann, H.-J., and Twes, F. J., "Lung Cancer Risk By Oral Exposure," in C. J. Bieva, Y. Courtois and M. Govaerts (eds.), Present and Future of Indoor Air Ouali _ty, Excerpta Medica, Amsterdam, pp. 91-100, 1989. Sakurai, R., "Prognosis of Female Patients with Pulmonary Tuberculosis," Jap~ Journal of Medicine, Vol. 28, No. 4, 1989. Samet, J. M., Humble, C. G, and Pathak, D. R., "Personal and Family History of Respiratory Disease and Lung Cancer Risk," Am Rev Respir Dis, 134:466-470. Schenker, M. B., Samet, J. M., and Speizer, F. E., "Risk Factors for Childhood Respiratory Disease: The Effect of Host Factors and Home Environmental Exposures," Am Rev Res'p_ ir Dis, 128:1038-1043, 1983. Schievelbein, H., "No Proof for Injuriousness to Health," Munch Med Wshchr, 124:1-2, 1982. Sheen, S. J., "Detection of Nicotine in Foods and Plant Materials," Journal of Food and 91

Science, 53:1572-1573, 1988. Shimizu, H., Morishita, M., Mizuno, K., Masuda, T., Ogura, Y., Santo, M., Nishimura, M., Kunishima, K., Karasawa, K., Nishiwaki, K., Yamamoto, M., I-Iisamichi, S., Tominaga, S., "A Case-Control Study of Lung Cancer in Nonsmoking Women," Tohoku J P.,~p_.M~, 154:389-397, 1988. Smith, M. L., "Publication Bias and Meta-Analysis," Evaluation in Edu~ti~;m, 4:22-24, 1980. Sobue, T., Suzuki, T., Nakayama, N., Inubushi, T., Matsuda, M., Doi, O., Mori, T., Furuse, K., Fukuoka, M.,Yasumitu, T., Kuwabara, O., Ichitani, M., Kurata, M.,Kuwabara, M., Nuashara, K., Endo, S., Hattori, S., "Association of Indoor Air Pollution and Passive Smoking with Lung Cancer in Osaka, Japan," Jpn J Cancer Clin, 329-333, 1990. Stahlberg, M-R., "The Influence of Form of Day Care on Occurrence of Acute Respiratory Tract Infections Among Young Children," Acta Pedlar, 282(S): 1-87, 1980. Sterling, T. D., Letter to the Editor, British Medical Journal, 282:1156, 1981. Strangert, K., "Respiratory nlness in Preschool Children with Different Forms of Day Care," Pediatrics, 57:191-196, 1976. Sutton, G.C., Letter to the Editor, British Medical Journal, 282:733, 1981. Svensson, C., Pershagen, G., Klominek, J., "Smoking and Passive Smoking in Relation to Lung Cancer in Women," ~.Q13f, gJg.g~ 28:623-629, 1989. Sydney, S., Caan, B. J. and Friedman, G. D., "Dietary Intake of Carotene In Nonsmokers With and Without Passive Smoking at Home," Am J Epidem 129:1305-1309, 1989. Tenkanen, L., Teppo, L., and Halmlinen, T., "Smoking And Cardiac Symptoms As Predictors Of Lung Cancer," J Chron Dis, 40:1121-1128, 1987. Tewes, F. J., Koo, L. C., Meisgen, T. J., Rylander, R., "Lung Cancer Risk and Mutagenieity of Tea," Environmental Re~e~r~:h, 52:23-33, 1990. Thuan, N. T. L., Migueres, M. L., Roche, D., "Elimination of Caffeine Interference in HPLC Determination of Urinary Nicotine and Cotinine," Clin Chem, 35:1456-1459, 1989. Trichopoulos, D., Kalandidi, A., Sparros, L., MacMahon, B., "Lung Cancer and Passive Smoking," International Journal of Canefr, 27:1-4, 1981. Tsokos, C.P., Letter to the Editor, British Medical Jou~al, 283:1464-1465, 1981. 92

U. S. EPA, Guidelines For Estimating Exposures, 51 Fed. Keg. 34042, September 24, 1986. U. S. EPA, Guidelines For The Health Risk Assessment Of Chemical Mixture, 51 Fed. Ike.g. 34014, September 24, 1986. U. S. EPA, The Risk Assessment Guidelil~es Of 1986, EPA/600/8-87/045, 51 Fed. ]~g. 33992, August, 1987. U. S. EPA, Workshov Reoort On EPA Guidelines For Carcinocen Risk Assessment: Use Of Human Evidence, EPA/625/3-90/017, September, 1989. U. S. Surgeon General CO. S. SG) (1964) Smoking and Health: Committee to the Surgeon General of the Public Health Service. Health, Education, and Welfare, Public Health Service. - Report of the Advisory U. S. Department of U. S. Surgeon General (U. S. SG) (1986) The Health Consequences of Involuntary Smoking. U. S. Department of Health and Human Services, Public Health Service. Varela, L. R., "Assessment of the Association Between Passive Smoking and Lung Cancer:." A Dissertation Presented to the Faculty of the Graduate School of Yale University in Candidacy for the .Degree of Doctor of Philosophy, May 1987, pp. 1-176. Wald, N. J. and Ritchie, C., "Validation of Studies on Lung Cancer in Nonsmokers Married to Smokers," Lancet, i:1067, 1984. Ware, J. H., Doekery, A., Spiro, A., Speizer, F. E., Fcrris, B. G., "Passive Smoking, Gas Cooking, and Respiratory Health of Children Living in Six Cities," Am Rev Respir Dis, 129:366-374, 1984. Willett, W. C., "Vitamin A and Lung Cancer," Nutrition Reviews, 48:201-211, 1990. Witorseh, R. J., Witorseh, P., "A Critical Analysis of the Relationship Between Parental Smoking and Pulmonary Performance in Children," Off Gesundh-Wes, 51:78-83, 1989. Witorsch, R. J., "Parental Smoking and Respiratory Health and Pulmonary Function in Children: A Review of the Literature and Suggestions for Future Research," in D. J. Ecobiehon and J. M. Wu, eds., Environmental Tobacco Sm0kf, Lexington Books, Lexington, MA, 1990, pp. 205-226. Wolf, F. M., "Meta-Analysis: Quantitative Methods for Research Synthesis." Beverly Hill, CA: Sage Publishing, 1986, pp. 1-65. Wu, A. H., Yu, M. C., Thomas, D. C., Pike, M. C., and Henderson, B. E., "Personal And 93

Family History Of Lung Disease As Risk Factors For Adenocarcinoma Of The Lung,= Cancer Research, 45:7279-7"284, 1955. Wu, A. H-T., "A Casv-Control Study of Lung Cancer in Women:" A Disse~afion Submitted in Partial Satisfaction of the Rvquiremcnts for the Degree Doctor of Philosophy in Public Health, 1983, pp. 1-192. Wynder, E. L., "Workshop on Guidelines to the Epidemiology of Weak Associations," Preventive Medicine, 16:139-141, 1987. Zeigler, R. G., Mason, T. J., Stemhagen, A., Hoover, R., Schoenberg, J. B., Gridley, G., Virgo, P. W., and Fraumeni, Jr., J. F., "Carotenoid Intake, Vegetables, And The Risk Of Lung Cancer Among White Men In New Jersey," Am J E_Didemiolo~y, 123:1080-1093, 1986. W#14804 94