Tobacco Documents Online

1 I I EPA COI4IEN'PS ON CRS DRA.FT "ENVIRONMENTAL TOBACCO SMOKE CANCER RISR" (June 5, 1995) Gaaraau co1oasarrs AND LUNG 1. The Draft has no stated purposa-and seems not to be dealing with any current or proposed legislation. Without a stated purpose, review comments become difficult. If the purpose is to provide a de novo CRS analysis of ETS and lung cancer hazard or risk, the analysis lacks rigor. If the purpose is to provide Congress with a better CRS review of the lung cancer section of the EPA Report (on the Respiratory Health $ffacts of Passive Smoking) than was done in the previous CRS report, then this should be stated. If the purpose is to provide Congress with an analysis of the OSF1A proposed Indoor Air Quality Rule on E'TS, then the Draft should have stated *this (and CRS should have provided comments to the OSAA docket). If ths purpose is to provide an uncritical analysis of tobacco industry vs. Govarn=ent/scientific and public health community positions, then this should be stated and arguments should be identified by source. Currently, -xquments presented by industry and their consultants are not properly identified. Without a stated purpose, we will assume that the purpose of this Draft is to do an independent analysis of ETS and lung cancer risk, giving arguments from both sides equal a priorl credence and relyinq on the author's analytical abilities to arrive at an unbiased judgement. 2. The Draft's literature rsview is inadequate, especially with respect to ETS exposure. For the exposura assesssant, too much weight is given to the unpublished Jenkins at ai. U.S. data and the 13 country Riboli et al. results, which include non-U.S. populations, while other published U.S. results are not reviewed. Faulty use of both these studies in the Draft's risk assessment leads to erroneous conclusions, especially about the relativs home vs. workplace 8TS sxposurss, and subsequent risk estimates. 3. The Draft's analysis of lung cancer risk, mostly limited to three recent studies, loses perspective by failinq to show the consistency of results in thirty other studies from different countries. Since bias is mostly a function of study design, and potential confounding factors for lung cancer are often cultural or dietary, failure to acknowledge these consistent increases from many studies in eight different countries could lead to the Draft's overstating the asount of uncertainty. For example, the amount of smoker sisclassification bias necessary to explain the observed increased lunq cancer risks in Japan and Greece is over 50%, far higher than has aver been hypothesized. 6 ra 1 I I I I I I I I I I I I I I I

I I I I I I I I I I I I I I I 4. The Draft takes the potential upward bias of smoker-status sisclassification as truth and as if it is the only operational bias, and then seeks to determine the amount of smoker status misclassification necessary to observe the increased lung cancer risk in one study. The Draft's method is improper for several reasons, detailed within our specific comments, and leads to the incorrect conclusion that "only a small number of case group members would need to be misclassified as never smokers to account for all the measured risk" (CRS-2). The Draft's approach in this area leads one to be suspicious of its intent (see also next point). 5. There is almost no consideration of the substantial downward bias on the lung cancer relative risk (RR) estimates from two types of exposure misclassification. The first type occurs, because people with nonsmokinq spouses are still exposed to other sources of BTS and there is no clean "unexposed" group. The second type occurs because some people whose spouses smoke may not actually be exposed to auch of the smoke; e.g., some spouses who smoke a lox say not smoke such at home. Analyses of lung cancer risk from workplace ETS exposure are generally prone to even greater exposure misclassification than the spousal smoking analyses. Failure to consider these well established downward biases for 8TS could lead to erroneous conclusions, especially ons of a possible threshold. 6. The quantitative risk assessment methodology used to estimate the lung cancer risk is seriously flawed. First, there is no sound rationale for the Draft assuming an effective threshold at exposure levels of practical concern. The reliance on portions of two studies as total support for an effective threshold, iqnorinq the exposure-rasponse results from other studies, several of which report increased risks from lower exposures (see EPA Report Table 5-11), suqqests, at worst, selective biased use of data. At best, the Draft ascribes too much sensitivity and precision to the low exposure region of the two epidemiology studies, especially qiven the exposure misclassification problems in all t$ess studies. Furthermore, the methodology used allows no margin for variability in the human populatioa or measurement error. In addition, the results presented rely on a faulty exposure assessment (see specific comments below). 7. Appendi= C - While the EPA has no official position on the risk of cardiovascular disease (CVD) from ETS, the CRS analysis presented in Appendix C is cospletely inad:quate. In addition, it is unclear why heart disease is beinq included in a report entitled "Environmental tobacco smoke and lung cancer risk". If the Draft is intended to be more comprehensive than a review of only lunq cancer, it should acknowledge that there are also other health effects attributable to ETS, including noncancer respiratory disorders, especially in children. w v 2 CO N . . ~ I

8PECI?IC COIO[EI1Tf OvQrviow 1. The EPA does not agree with the suggestion that its approach to assessing the lung cancer risk in its 1992 Passive Smoking Repoit "may be misleading" or the claim that "it appears that any risk which aay be present is concentrated in a relatively small portion of that (the adult nonsmoking] population" (C1tS- 1). The EPA's approach was reviewed in public session by independent scientific experts in the field, who concurred in the EP71's methodology (See attached SAB Report). The EPA had already sent these SAB comments to the CRS, along with the EPA's review of the previous CRS report. A responsible CR.S approach would be to acknowledge the degree of concurrence the EPA report has received, rather than to simply pejoratively state that "it may be misleading". 2. Furthermore, there is no support for the Draft's claim that the rasainder (not exposed to ETS at home) of the nonsmoking population would be "at a minisua, 20 times less likely to be at risk" (CRS-1). This conclusion, based on looking at parts of two lung cancer studies and one exposure study instead of the whole body of evidence suggests a conclusion in search ait data. This selection of data is exactly what, on the very next page, the Draft accuses OSHA of doing, "OSHA based its risk assessment on one such estimate, which indicated an increased risk, and disregarded the remaining estimates which found no overall association between workplace exposure and lung cancer." In short, the methodology used by the Draft to generate its lung cancer death estimates is seriously flawed (CRS-i). (See our specific comaents on the dose-response section). 3. The overview of "sources of uncertainty" (CRS-2) omits exposure misclassification, a major sourc. of downward bias. The EPA agrees with the conclusion that potential confounders cannot explain the observed increases in RR. However, the Agency does not support the conclusion that smoker misclassification bias could be responsible. The EPA Report actually corrects each of thirty individual studies for potential smoker-status misclassification bias bsL= doing specific analyses. (See other coments on dose-response section). 4. Regarding the overview of "occupational risk• (CRS-2), the Draft's conclusion that "residential and other non-workplace exposure tends to exceed workplace exposure by a factor of 4 to 6" is based on faulty comparisons in the Jenkins at al study and is incorrect. A proper analysis of that study will reveal that workplace ETS levels where there are no smoking restrictions'are higher on average, as well as for 80th and 3 I I I I I ~ I I I I I I I I I I I

I I I I I ~ 95th percantilss, than are smoker-occupied hoae ETS levels. Even if workplace ETS exposures were on averag. slightly lowar than smoker-occupied residential exposures, this does not preclude sany exposed workers from being at increased lung cancer risk. 5. Regarding the overview of "biological and chemical issuss" (CRS-2), difficult, although measurements of ETS concentration are the EPA concluded that nicotine is generally a good marker for ETS exposure (1992 BEPA report, Chapter 3). 6. The concluding statement (CRS-3) that "thara raaains considerable uncertainty about the biological processes linking ETS to possible onset of lung cancer" is specious; the same claim could be made of active smoking or asbestos, yet - there is no doubt that they cause lung cancer. , Introduction , 1. The HPA, objects to tha.Draft's claim (CRS-5) that "whil• (th.  EPA report] has received support from the public health community, it has been criticized on scientific grounds". In fact, the EPA report has received support from the public ~ health community and from the larger scientific coaiunitv based on its scientific marit. The Draft should acknowlsdqs that the EPA's conclusions are very similar to those of the ' 1986 report of ths National Research Council of the National , Academy of Scianc.s, a non-qovarnmental body of the Nations most prestigious scientists, certainly part of the scientific community. Similar conclusions in both the 1986 Surgeon ~ Gane.ral ls and the 1991 NIOSH Reports have been authored by scientific axp.rts and received extensive review before release. Also, the National Cancer Institute has republished the EPA Report in slightly abridged form in a monograph series ~ under its own logo. By sharp contrast, criticisms of the EPA Report have qenarally bs.n orchestrated by thee tobacco industry and have not withstood scientific scrutiny. The I Draft's oquatinq industry supported criticism of• these critically, reviewed assessments with the assessments themselves suggests a bias in ths Draft"s development. ~ , 2. '  ~  I Furthermore, the assessments are all based on a substantial body_of literature, a fact which the Draft ssams to disregard. Th.+pr sviQu~, CRS report, cited on CRS-5 as alsq being critical of 3;':'F~~s analysis, was seriously flawed, aod any citation of it should be accompanied by a refarenoe to EPA's rebuttal (See attached letter to Dan Mulhollan). Isn't that why.th, current Draft is being prepared? Furthermore, the statement in the footnote that, "It was necessary (for the CRS) to review the evidence of a passive smoking health risk because this is a potential component of the cost calculation" is open to interpretation. As an economic analysis, the previous CRS ~ 4 "~ aa ~ ro ~ ~ CO

report by Gravalla and Zimmerman went beyond its expertise and its scop. in reviewing the "evidsnca of a passive smoking health risk". Since the potential "external" or "spillover" costs of health effects of ETS are so small compared to the direct costs of health effects from active smoking, it was "necessary" only to put a cost value on these "axternals", not to discuss the uncertainties as the authors saw thes. In fact, the authors of the previous CRS Report on Cigarette Taxes discuss fairly extensively the "evidence on passive smoking [hsalth] effects", but they never discuss the evidence on the health effects from activa!smokinq, which contribute far more to the health costs. The question of "why?" is, we feel, connected to the question of "what is the purpose of this draft?" 3. The statament that "confounding and smoker misclassification (ars) the two principle sources of uncertainty in the epi studias" (CRS-6) again iqnoraa the major downward bias resulting from exposure misclassification in these studies. 4. On CRS-7, footnote 10, OSHA is part of the Dept. of Labor, not Energy. The same mistake occurs in footnote 48. Spideaiologic Studies ' 1. On CRS-8, paragraph 3, the Draft reads as if the EPA adjusted the risk estimates from the individual epidemiology studies for background ETS exposure. This needs to be clarified. The EPA did adjust the individual risk estimates for potential smoker misclassification bias in its hazard idsntification analysis; however, it did not adjust the individual risk estimates for "background" bTS wrporurs. uter conducting its hazard identification analysis, the EPA did adjust the pooled overall O. S. RR estimate and the overall Pontham at al.(1991) RR estimate for EPA's two quantjtativs risk analyses. 2. In the section on bias and confounding (CRS-11), it should be noted that bias and confounding can work in both directions, and that nondifferential biases exert a downward bias on the relative risk estimate. 3. The statement (CRS-12) that "EPA adjusted the risk•astimates to account-.,for bias and confoundinq ..." is incorrect. EPA did axaminr•potantial bias and confounding and concluded that they could not account for the consistent dosa-rslated increases in lunq cancar risk observed across numerous studias from different countries. However, the only adjustment that EPA mada to the individual relative risk estimates was for the potential upward bias from smoker misclassification. There is no consistent evidence that any other upward bias is operative in thasa apidemioloqic studies, and even the evidence for an 5 I I I I I I I I I I I I I I I I r3o ' ~ oo _cz I

I I I I I I I I I I I I i I I I I upward bias from smoker misclassification is not conclusive (Saa discussion of smoker misclassification below). 4. In discussing the consistently increased risks in the highest exposure groups (CRS-12 to CRS-13), it is misleading to single out the U.S. studies and contend that only two of these six were statistically significant without considering the issue of the low statistical power in most of thesa epidemioloqic studies, especially when the study population is broken down into smaller exposure subgroups. 5. In the section on meta-analysis (CRS-13 to CR3-ls), it should be clarified that the meta-analysis of pooling overall relative risk estimates by country was only one of several analyses performed by EPA and that EPA's conclusions do not rely on the meta-analysis. 6. •In the paragraph (CRS-13) discussing the EPA Report's weighting of the relative risk estimates, the Draft's wording is very deceptive. it"implies that EPA did something devious by giving the larger studies more weight. In fact, EPA weighted the studias by the inverse of the variance, which is stan$ard statistical methodology. The same deceptive wordinq was used in the previous CR8 report and was addressed in RPA's rebuttal (Sea attached letter to Dan Kulhollan, section D.3.). 7. In discussing the pooled relative risk estimate of 1.19 in the EPA Report from the U.S. data (CRS-13), it should be emphasized that this is ths pooled estimate for the effects of spousal smoking, unadjusted for other sources of ETS exposure. 8. The discussion of $PA's use of a one-tailed significance test is very misleading and slanderous (CRS-13 to CRS-14). A one- tailed test is a standard statistical procedure when there is prior evidence that if there is an affect, it is likely to be in a specific direction. EPA did not "(argue] that there is no biological reason to expect passive smoking to reduce the risk of lunq cancar• (CRS-14) but, rather, demonstrated that based on prior knowledge of the carcinogenicity of active smokinq and the similarities between mainstream saoke and ETS, any lung cancer affact of BTS would be expected to be causativa versus protactive. In addition, with respect to the "crucial point• (CRS-14),, where EPA is accused of changinq tests between drafts, the CRS draft is totally erroneous; EPA used a one-tailad test in both drafts as well as in the final report. Bnvironmentsl Tobacco 3'soke 1. There are more than 40 known or suspected human carcinogens in STS not just 14 as mentioned on CRS-17. 6 I

I 2. The CR3 draft correctly concludes.(CR,4-lg) that "cigarette- equivalent exposure estimates vary siqnificantly depending on which STS component is chosen" and "are probably of liait.d use in determining risk". Then, three pages later (CRS-22), the draft contradicts itself and inappropriately suggests that "typical ETS exposure is equivalent to saokinq approximately one-tenth of a cigarette every day", based on relative nicotine levels. 3. The near total reliance on the unpublished Jenkins et al. study to deteraine ETS exposure represents, at best, a highly selective and questionable use of available data. This extensive dependence on a tobacco industry-supported study, with tobacco industry co-investigators, hardly gives assurance to EPA and, we suspect, to a great many research scientists, as well as the public health community, that the CRS, with all its literature retrieval resources, is interested in an unbiased analysis. 4. Furthermore the Draftls analysis of the Jenkins .t al. data is incorrect. Table 3(tR.S-21) is used to show the "average nicotine concentration and actual nicotine exposure", based on the recent RJ Reynolds - Oak Ridge (RJR-OR) unpublished 1i city U.S. study. The implication is that "rssidential anil other non-workplace exposure •tends to exceed workplace exposure by a factor of 4 to 6" (CRS-2) and (CRS-47), so "that many workers would not be exposed to sufficient ETS to be at increased risk for lung caAcer" (CRS-2). However, this interpretation is incorrect. The correct conclusion to be derived from results in the RJR-Oak Ridge study is that where there are no ssaking restrict3ons, 8TS coac.ntrations at boass and t n the workplaco are siAilar. In soae workplaces, 8TS levels and average daily exposures are considerably higher than at hose. This is saen in the enclosed Table which compares the RJR-OR nicotine concentrations in "hoss" and "work" settings. The Table also presents estimates of total daily exposures based on two breathing rat. scenarios: 1) 1.2 al/hr. .( us.d by CRS ), and 2) 10 s3/ 8 hr. at work and 10 as/ 16 hr. at hoa. (sore physiologically realistic). Caaparison of the two tables highlights several flaws in CRS' Tablo 3. First, the "non-work" n-548 should not be used as it includes the n-41s hose and is ;rrdundant and misleading. Second, the home (n-t1s smoker occupied- hoses) should be compared with the work n-168 (no smoking restrictions). This allows direct comparisons in unrestricted saokinq environments. Third, the exposure comparisons should be made using the breathinq-rate scenario of 10 m~ for both hose and workplace (this is what both EPA and OSHk use). Also, weekend non-work conditions should probably be taken into account. I I I I I I I I I I I I i I I ~ 7 IM ~ ~ ~. I

ar .. w. w.. .r r i.. .r r+.. r.. .M s+r r a.. en Table: B'1'S Nicotine C.awaUrations wrd Exposure Pstimata for Smoher-oocupiod Homes versus Wodcplaoes Based on Jenkuu et al Data and Various Breadtin,8ltate Aswmptioas' BNYIRONM6NT NICOTINE AIR CONCENTRATION (uWm') EXFOSURE (ui/d.y) at 1.2 m'/hr .t 10 m'/day 6ome ar waic Home 16 rr MwA at 15 hr x 1.2 m'/hrr st 19 m'/16 hr (N •• 4 15) (UN*W-OMpW boew mly) me" . 0.68 12.1 6.8 mm 2.16 38.8 21.6 BQq~ ~k 2.79 48.7 27.9 ~ 7.10 127 71.0 Wat ~brTwA at t ar x 1.2 m'/hr at 10 ®'/8 br (N w 520 Toad No Smokiq Ratriction 1'Lu Sa~oki~ 1~ ~ No S~ ~~ $~ ~ no*aod A~• t?~I~/) " p~~+ R,~pMiriotio~l DodVMod Dedjn~led No Smoking ~~ Atrn Oldy No Snoidq6 SaWmilim Ann Only Me S Ratricxlons N s 168 ,no" 0.09 0.9 a,P 5.6 0.9 5.8 : Muuu 0.79 2.77 7.5 26.6 7.9 27.7 IiDt6 por+caWZo • 0.48 3.59 ' 4.6 34.5 4.8 35.9 9Sth porawdle 3.90 14.0 37.4 134.4 39.0 140.0 $Re( - SAB CRS: 'ttibb 3: Rqmura .Ad Aitbaps ~x~oeatrukr~ of NbodAO (U.S. )6{.:'!y Study) of Iodividu.lx Reoordft Tobacao Pioduct ' ' IUb1a 6, 7, tW t. l=WMy 1V4?J94, Ob~lb~ in Airlwa Vam (QIB. CRS-21). RM bM UMMS . OSHA wlb4 • 12.4% Tcol Noatibor ot WakoM SaopMd 1336 z6COBUtoz

i In addition, note that only 12.4% (168/1356) of tha total numbar of workplaces samplad by Jenkins at al. had no smoking restrictions. our recollection is that national surveys suggest that at least 358 of all U.S. workplaces have no smoking restrictions. In any event, workplace HTS exposures are far aore variable than are homa axposur.s. This ia because of the high variability in workplacs sizes, ventilation rates and smoker densities. This is sosawhat apparent in the coaparisons presented in the enclosed Table. Note that while the aadian cotinina concentrations are higher for hom: vs. work (0.68 vs. 0.58), the corresponding maans (2.16 vs. 2.77), 80th percentiles (2.79 vs. 3.58) and 95th percentiles (7.10 vs. 14.0) are all lower. Thus, even a blanket statasent of comparability of home vs. workplace exposures is too precise. Again, the CRS has depended mainly on the RJR-OR study for its exposure aatiaatas, and has mada the wrong comparisons. Thara are many mora (and published) studies availabla.. The CRS sathodoloqy and primary dependence on this unpublished study for a conclusion of low workplace axposura should be corrected. 5. with respect to the discussion of biomarkars (CRS-22), a compound need not be directly related to a specific health risk to be a markar of 8TS exposure. DosQ-Rospons• Relationships 1. The EPA Report did not "(assuma] that the entire population of never saokars was subject to the saas risk" (CRS-24). Rather, the Report used av.rage, risks, as aatimatad by the apidamiology studies, to calculate population astisatas, and different average risks wara used for people with and without ETS exposure fros spousal smokinq. An alternative could have been to usa sos. sort of linaar dosa-basad approach, although there is substantial uncertainty regarding what doses are associated with what exposures. However, to infer a threshold froa these apidasioloqy studias, as the Draft has done, is not justifiable, especially given the exposure misclassification problams inherent in these studies. 2. On CRS-26, the increased risks observed in the lower exposure groups of the Stockwall at al. study should not be dismiss.d as "not statisticallx siqnificant at the 95% confidanca laval.• These small exposure subgroups have low statistical power and 8'TS is a known human carcinoqanf therefore, the observed increased risks in these lower exposure groups are likely to be biologically significant, avan if not statistically significant. 0 Ct 4h 8 ~ ~ ~ w ~ I I I I I I I I I I I I I I I I I

I I ~ 3. Also on CRS-26, it is totally inaccurate to determine that "the risk values at the lower exposure lavals are all sufficiently close to unity to conclude that there is a high ~ probability that only the highest exposures to ETS aay present any risk of developing lung cancer". These studies of spousal smoking are subject to substantial exposure misclassification becauss many people are exposed to ETS from a variety of . sources, not just spouses, and there is no truly "unexposed" ! comparison group. The effects of axposure.aisclassification would be especially pronounced in a low spousal exposure _ group, because thase women with low spousal exposures and I variable other exposures would be compared with women without spousal exposure but also with variable other exposures. The high background "noise" from the variable other exposures in ~ both groups is highly likely to mask any sffact of the low spousal oxposure! 4. The CRS draft ascribes far too auch precision to the relative ~ risks reported for individual subgroups. These are relative risk eitisatea, based on a crude exposure surrogate, not a true indicator of total dose; and observed in a small , subsp.apla of the population. Overall, the study results show consistent dosa-related increases in lung cancer risk from ETS exposure. However, the specific estimates should not ba treated as exact, and the C,RS should not dwell so auch on an I apparent anomaly in a specific dose-response relationship (e.g., discussion of Fonthaa et al. results based on total years of exposure, CRS-26 to CRS-27). / 5. Regarding smoker aisElassification bias (CRS-31 to CRS-35}, an a inherent but unspoken assumption is that former or current , ! ~ ' , _  w 6. , / ! sackers are equally likely to be aisclassified as nevar- saokers whether or not their spouses saoka. Then, it is explicitly assumed that aisclassified sackers are more likely to be married to other smokers because smokers tend to marry other smokers. Howevar, there is no evidence that the first assumption is accurate. In fact, it seems plausible that smokers married to other sackers may be less likely to misrepresent th.aselves as nevar-smokars in self-reports than sackers married to nonsmokers, who aay faal aore sslf- conscious and stigmatized about their smoking habit. If this is the case, then smoker misclassification may not actually result in any measurable upward bias in thase apideaiologic studias,~ and may even produce a downward bias. The CRS analysis of the smoker aisclassification rate required to eliminate trend in the Fonthaa et al. study (CRS-32 to CRS- 35) relies on an unstated assumption that there is no saoker misclassification of cases in the baseline comparison group (i.e., 0 pack-years exposure), vhich may not be warranted. It apparently also assumes that there is no misclassification in the control groups, although since.thara are,no accompanying 0 9 ~' ~ ~ .~ ~ ~ .~ I

calculations, we couldn't confirm these fiqures, At the very least, all these assumptions should be made explicit. 7. The Fontham et al. study was designed specifically to minimize potential smoker-status misclassification bias. Kisclassification rates are reported in their paper and strongly suggest that in their study design, there is no upward bias due to smoker status misclassification. In fact they suggest, if anythinq, a small downward bias in their study. Thus, Table 5 ignores the available data in order to hypothesize a "what if" scenario. 8. With respect to the "second question" raised at the bottom of CRS-33, i.e., "whether there is reason to believe that the (saoker] misclassification rate would be higher for those most intensively exposed", such speculation is counterintuitive. 3mokinq women whose spouses snoke a lot are more likely to smoke a lot themselves and less likely to be misclassified as never-ssokers. It is generally former smokers and occasional current smokers that are misclassified (about 20), not heavier current smokers (about lt) (See EPA Report, Appendix B) . _ 9. Also, see again the comments above on smoker-status misclassification rates necessary to account for the observed increasas in Japan and Greece (General Comments #3). BTS and Lung Cancer Deaths 1. On CRS-35, "backqround" ETS exposure needs to be defined so that it is clear that this refers to all sources of exposure (e.g., work, social settings, etc.) other than the spouse. 2. On CRS-36, in discussinq the approach used by EPA to estimate the population risk, it should be noted that EPA used two approaches, one based on the pooled estimate for the II.S. studies and the other based on the results of the Fontham et al. (1991) study. 3. On CRS-37, CSS claims that "in essence, background ETS, accordinq.to the =PA and NRC calculations at least, appears to be considerably aoree potent than spousal BTS". The difference in.amount of risk from spousal versus background ETS reflects differences in amount of exposure, not potency. Claiming one source of ITS is more- "potent" than another suQQests that there are differences in risk from the same amounts of ETS generated from different sourcea. Furthermore, som. of EPA's calculations (based on the Fonthas et al. study; See EPA report, page 6-23) found the same risk or greater risk from spousal exposure as from background STS. Similarly, the NRC calculations estimate more risk from spousal than from I I I I I I I I I I I I I I I I r~ - 10 ~ ~ . ~ ~ ~ I

I I I . I I I 5. I I 6. I I I I . I I I 8. I I I background sources of ETS, contrary to what the CRS draft suggests. Also on CRS-37, the CRS draft claims that "EPA assumes that atter a certain point, the risk of ETS becomes greater to these individuals than the risk due to having smoked". This statement is incorrect. The ETS risk to tormer smokers is not necessarily greater than the risk from having smoked, but it is greater than if they were not involuntarily exposed and it becomes relatively more significant compared to the "voluntary" risk they assumed from previously smoking. On CRS-38, CRS claims that "the (LCMRJ method just described is exactly equivalent to assuming that the values of relative risk... are the sase...". This statement is false. ' As groups, both former smokers and male never-smokers have greater background risks of lung cancer than female never- smokers; therefore, it.they have the same ETS exposures and risks (LClQts) as femala never-smokers, they would have lower, not the same, relative risks. On CRS-38, it is illogical to claim that SP71 assumed that all of the never and long ago former smokers who have been expos?~si to spousal smoking have been subjected, on a.eraqe, to ths same combined level of spousal and background sTS, and all never and long ago former smokers not exposed to spousal ETS, received, on averaQe, the same amount of background ETS (emphasis added)". If the point is that never and former smokers were assumed to be exposed to th. same average Lveli, then the issue of dose-response in the remainder of the paragraph does not follow. It appears, rather, that CRS is trying to suggest that EPA assumed everyone was exposed to the same level rather than an average level (See also comment #1 in the Dose-Response Relationships section). Also on CRS-38, CRS states "that the risk only becomes maasurable after a long term, intense exposure to ETS (emphasis added)". "lieasurable" is a critical concept given the imprecise nature of the epidemioloqic studies and the absence of a truly "unexposed" group (See comment #3 in the Dose-Response Relationships section). CRS then ignores these issues and assumes that there is no risk at exposure levels below those for which the risk becomes measurable. CRS tbenAclaims that fthe number of non smokers ... who are' exposed to levels of ETS... extensive enough to be at a significant risk of developing lung cancer is likely to be considerably smaller than that entire non smoking population (as was assumed by EPA and NRC)" (CRS-38). cRS does not define "significant risk". If EPA had done a dose-based quantitative risk assessment, people exposed to lower than average ETS levels would, in general, have lower than average 11

I risks; however, this would be offset by people with above average exposures having above average risks. 9. The suggestion that "the 13!" with the highest ETS exposure is at "at least 20 times" greater risk based on coaparinq unadjusted RRs of 1.42 and 1.02 from two different exposure levels (CRS-39) again ascribes too much precision to the specific rasulta of a single epidasioloqic study. The comparison also suffers from the use of the unadjusted RRs rather than the adjusted RRs. 10. Furthermore, the 13% of the Pontham et al. controls exposed to 40 or more pack-years of spousal smoking may not represent the entire adult U.S. nevarsmoking population. The controls were selected to match the cases on aqe and race; they ware not selected as a random population sample. 11. The Riboli at al. 10-country results may not reflect exposures in the o.S. . 12. On CRS-40, paragraph 1, it is unclear why CRS switches from discussing the "13:" exposed to 40 or more pack-years (CitS-39)- to the "4%" exposed to 80 or more. It does not follow that "it is probable that only 2 percent of the background only group was subjected to $TS' exposure amounting to the equivalent of 40 pack-years"; half of 13% would be 6.5% by these calculations. 13. On CRS-40, paragraph 3, CR4 repeats its mistaken assertion that the LC'!at method "is equivalent to assuminq that the relative risks from ETS for [aale never saokers and long ago former smokersj is exactly the same as for female never smokers". See comment #5 above. 14. The "threshold" methodology proposed on CRS-41 for estimating lunq cancer deaths from ETS is seriously flawed. As described above, the apidemioloqic studies are not sufficiently pracise and suffar from exposurs misclassification; therefore, there is no sound foundation for assuminq that a threshold (or abruptly sublin.ar exposura-rasponsa curve) exists for exposure levels of practical concern (See comment #3 in the Dose-Response Ralationships'section). 15. In addition, thsre is no sensible scientific basis for assuming a"thrashold"-for a carcinogenic agent that has both initiating and promoting capabilities and has demonstrated its carcinogenic potential at typical environmental exposure levals. The concept of ascribing a practical threshold to low levels of exposure say have some merit when extrapolating from, say, high occupational to low environmental levels of exposure, which typically range over a factor of 100 to 1,000 times. It is much sor* problematical when the factor from 12 I I I I I I I I I I I I I I I I I

a I ~ ' . , 16. ~ ~ i  ' , / . ' - -! . r , 17. !   lower to higher is 2 to 3 times, and there is a damonstrated downward bias operating in addition. While the Draft does try to make that argument, we feel that it lacks logic. Considerinq how quick the Draft is to adopt uncertainties regarding unspecified confounders (especially for heart disease and ETS) as a rationale for rejecting observed associations, it is hard to understand how the threshold concept becomes so uncritically embraced. Furthermore, even if a"threshold" were assumed to exist, the methodology used by CRS leaves no marqin for human variability or saasuremant error. For exaaple, applying the EPA method for deriving reference doses (i.s., an estimate of the daily exposure to the human population, including sensitive subgroups, that is likely to be without an appreciable risk of deleterious effects during a lifetime; see attached reference by Barnes and Dourson) for noncarcinoqenic agents, which are presumed to have exposure thresholds for inducing noncarcinoqanic effects, to the exposurs-response results based on pack-years of spousal exposure from the lonthas et al. study miqht proceed as follows: The 15.1-39.9 pack-year exposure group (OR-1.02) siqht be considered a"no observable adverse effect leval•, or NOARL, while the 40.0-79.9 (OR-1.34) would be considered a"lowest observable adv~ rs4 effect level", or i.ZA8L. These levels are sensitive to the specific cut-points u..d and to the statistical power of the study to detect an effect in the exposure groups.- For example, a significant increase aay have be.a observable in the lower exposure groups if the sample size had been greater or in a 30-50 pack-year group if different cut-points had been used. In the absence of a NOAZL, the LOAEL is used with an uncertainty factor of 10, because of the uncertainty as to where the threshold may lie. In this study, the NOaEL of 15.1-39.9 would be usQd. Then an uncertainty factor of 10 would be applied for interindividual variability within the human population, to allo;r a margin of safety to protect more sensitive subpopulations. Finally, a sodifyinq factor of 2-10 siqht be applied to take into account the fact that lung cancer is a serious, generally fatal, health effect, as oppos.d to moderate changes in body weiqht, for exaaple. So, even if a threshold were assumed (which is not appropriate in the cass of ETSI), a sore responsible analysis would not assume that a"safe" axposure level exists abovs that which results in the dose received by the 15-40 pack-year group divided by 20-1001 - If the purpose of the CRS quantitative risk calculations is to distinguish between spousal and nonspousal exposures, the lung cancer deaths in nonsmokers exposed to spousal smoking should be broken down into thoss LCD's attributable to the spousal saokinq and those attributable to other sources of ETS (CRS-42 to CRS-43). As it is now, the Table (CRS-42) could be N CZ1 13 ~ t•.~ ~ ~ ~ I

misinterpreted as suggesting that all the LCD,,s in the spousally-exposed group are from spousal smoking. Occupat.ional S'TS Lunq Cancer JUsk 1. On C8S-43, CRS claims, with respect to quantitative risk assessment, that "EPA sade no attempt to assess the lung cancer risk from occupational (i.e., workplace) exposure to ETS, arguing that there are too few workplace ETS studies to conduct a aeta-analysis, and that it is difficult to obtain reliable assessments of workplace ETS exposure". In fact, EPA did not quantitatively assess the lung cancer risk from occupational exposure to ETS because it was beyond the scope of EPA's risk assessment. EPA does believe that the workplace lung cancer results are generally unreliable and should not be used in a aata-analysis. If EPA had done a workplace risk assessaant, the Agency most likely would have used the Fonthaa st al. occupational results, because the Fonthaa .t al. study is th. highest quality.study and it is a large multi-regional U.S. study. EPA did use the Fonthaa et al. (1991) results for one of its population risk calculations. 2. On CRS-44, last paragraph, it is not legitimate to claim that "only one of these was well-conducted, according to EPA". Spl1 categorized the biqhst quality (for th. purposes of obtaining information on BTS) studies in "Tier 1"i however, EPA's categorization does not assert that studies in lower tiers ware not "w.ll-conductul". 3. On CRS-45, with respect to the Srownson et ai. occupational results, it is worth noting that lung cancer risks were increased in the two highest exposure quartiles. 4. On CRS-45, CRS cites the results of the Lavois and Layard workplace aeta-analysis. There are, however, serious probleas with this analysis. Please sea the attached paper, "Passive smoking and lung cancer: the U.S. EPVs weight-ot-evidenca analysis, with emphasis on the epidemiology studies•. 5. On CRS-47, the Riboli et al. results aay, not be applicable to the U.S. 6. The statements on CRS-47 starting with "Stropq evidence ----" and anding with "These data suggest that the+workplace is not the dominant place of exposure and that workplace and residential exposures aay not be comparable" ar. all, at best, insupportable by the Draft's analysis of the Jenkins study, as detailed above. We suggest that a such better exposure analysis is needed, not only of the Jenkins et al study; but of the many other exposure studies. See also the coaaents above on exposure. 14 I I I I I I I I I I I I I I I I I

a I I I I I I I I I I I I I I I I I 7. The conclusion that "I!, on averaqe, workplace ETS exposure is lower than residential exposure, then it is likely that aost, if not all, workers would not be exposed to sufficient levels of bTS to be at increased risk tor lung cancer" is not warranted in any event. Even if high exposure levels were required to develop lung cancer, and even if avarage workplace levels were lower than average residential levels, there are a substantial nusber of workers, especially in the hospitality industries and in other workplaces with no smoking restrictions, that are exposed to very high levels of ETS. 8. The linal sentence, claiainq "The evidence ... casts doubt on OSHA's risk assessment and suggests that it say have siqnilicantly overestisated the occupational 8T3 lung cancer risk", is just not supported by the Dralt"s analysis. Apportdiz A 1. This presentation oi the EPA's conclusions about the respiratory health effects of passive smoking should include the Agency's conclusion that BTS is a known huaan carcinoqen, because this classification conveys the co:.clusion of a causrl, association and the strength of the evidence supporting this conclusion. Appendix B 1. Appendix B is Aisleadinq. It should state aore explicitly and immediately that this extrapolation illustrated lroa active to passive smoking pertains to the cigarette equivalent approach, and it should clarify that this was not the sethodology employ.d by the =PJ1. 2. The illustration ostits the possibility of a supralinsar dose- response relationship, which, incidentally, is likely to be the appropriate extrapolation lroa high to low doses of tobacco ssoks for cardiovascular risk (See coaunts on Appendix C below). Appendix C 1. The stateaant (pq. cR8-51) that "Whereas the EPA estisate of passive smoking lung cancer deaths used a value of Z-1.75, Wells ahose a higher value of Z-2.6, based on Fonthas"s study ..." is inaccurate. The EPA derived its estiaate using two different approaches, one of which eaployed Z values of 2.0 and 3.6 obtained troa the Fonthas study (See the EPA report pages 6-21 to 6-26). 2. The coaparison of an active smoking h.art disease risk of 1.7 with'a passive saokinq heart disease risk of 1.22 is flawed by 15 I

oversimplification. For one thinq,'studias of active smoking use navar-s=okars (including those exposed to ETS) as the baseline camparison group, while studies of STS use "unaxposad" navar-smokars as the comparison group; therefore, these relative risks are not directly comparable. In addition, because heart disease risk drops dramatically upon smoking cessation, a more raasonable comparison would be to compare the risks for current active smokers to those for current passive ssokers. Furthermore, the summary age- adjusted relative risk of 1.7 for heart dis.asa from ever smoking obscures the much larger aqe-specific relative risks from smoking observed in younger aqa groups (Saa, a.q., The B.alth Consequenc.s of 8moklnq:. Card.iovaicular Dissse, U.S. DIDiS, 1983, pqs. 103 and 113-115, and the recent study of Parish at al., 1995, Bi6T 311:471-7, which reports risk ratios for nonfatal myocardial infarction for current saokers vs. nonsmokers of 6.3 for 30-39 year-olds, 4.7 for 40-49, 3.1 for 50-59, 2.5 for 60-69, and 1.9 for 70-79.). Also, more recent studies of active saol4inq are reporting higher RRs for heart disease than earlier studias, apparently as a result of decreasing baseline risks for heart disease dua to recent lifestyle changes such as better diets and exercise. If these issues wara taken into account, the diffarence in RRs betwaeS active and passive smoking wouid be larger. 3. Furthermore, the CRS analysis does not satisfactorily present the possibility that a supralinear dose-response relationship exists between active and passive saokers with respect to CVD risk, i.e., that some of the effects of tobacco smoke on the cardiovascular system saturate at higher doses. The CRS draft does not discuss the mechanistic evidance supporting this hypothesis. For example, platelet activation appears to saturate at low doses of tobacco smoke, and nonsmokers appear to be more sensitive than smokers to tobacco smoke-induced andothelial damaqa (S.a Glants and Paralay, 1995, JA1i71 273:1047-1053). See also the recent studies of willett at al. (1987, N8J1t 317:1303-1309) and Palmer at al. (1989, NEJK 320:1569-1573), which both report statistically significant relative risks of about 2.4 in the groups that smoked only 1-4 ciq/day. 4. The CRS draft should discuss more fully the mechanistic •videnca that ZTS causes cardiovascular ef~ects.(See Glants and Parsley, 1995). For sxampla,'thars is strong avidence• that ETS promotes the development of atherosclerosis, a major cause of heart disease. Animali models have shovn that subchronic exposures to low levels of ETS promote the development of atherosclerosis, and there is human evidence that ETS exposure increases the carotid artary vall thickness, suggesting that 8'TS also contributes to the development of atherosclerosis in humans.- The role of ETS in atheroqanesis r.~ I I I I I I I I I I I I I I I I I _'

I I I I I r r 5. ., , i 6. ' ,  , ~ ~ ~ ~ ,  ~ , I is likely mediated, at least in part, by polycyclic aromatic hydrocarbons (PkBs). A recent biomarkar study suggests that passive smokers receive up to 40% of the PA8 dose of active smokers (Crawford at al., 1994, JNCI 86:1398-1402). The sachanistic evidence that ETS can create an isbalance.b.twean oxyqin supply and desand, as revealed by decreased exercise tolerance, and can activate platelets similarly warrants sore discussion. Other machanistic evidence is oiitted from the CRS draft altogether (See Glants and Parsley, 1995). For axample, anisal studies have deaonstrated that STS decreases the ability of the styocardius to process oxygen to produce ATP for energy and that ETS exacerbates ischeaia/reparfusion injury. The CRS draft suggests that "one cannot rule out the possibility that the heart disease risk values in ths api studies may be largely due to confoundinq", but presents no evidence that confoundinq is affecting the apidemioloqy results. The draft claiss that nonsmokinq spouses of smokars are likely to share soma of the unhealthy lifestyle characteristics -qenarally associated with their ssoking spouaas, but again presents no evidence that this explains any of the epide'ioloqy results. Zn the discussion of contoundinq in the lung cancer section (paqes CRS-27 to CRB-31), it is noted that Le Marchand (1991) reported decreasing cholesterol and total fat consuaption with incraasinq ETS exposure, while Alavanja at al. (1993) found no correlation between saturated fat intake and exposure to ETS. Fonthaa et al. (1994) reported an increased risk of lung cancer associated with dietary cholestarol, but their lung cancer results from sTS exposure were not affected by dietary cholesterol, similarly suqqestinq no correlation between dietary cholesterol intake and exposure to STS.' Both cholesterol and fat consuicption are risk factors for CVD. The CRS draft also cites. Friedman at al. (1963) as raportinq a correlation between 8TS exposure and alcohol consumption. Moderate alcohol consusption is generally considered to be protective aqainst CVD. These results, presented by the CRS, suqqest that these "potential confoundinq factors" ari either not associated with ETS exposure or say actually be exerting a downward blas on the relative risk estimates for CVD. The CRS draft notes that only 4 of 12 epidasioloqy studies controlled for at least 6 potential confounders and that the largest study failed to control for any of thes (although it did, in fact, attempt to control for SES by adjusting for education and quality of housing). Six is not a maqic nusbar, and not all of the potential confounders are equally important. Roughly half of the itudies controlled for at least the major CVD risk factors of hypertension, cholesterol, ~~ ~ 17 ~` Co Pr ~

I l and obesity. In addition, Wells (1994) observed that the relative risk estiaates of studies that adjusted for the aost factors were generally higher, indicating that the net effect of any confounding in these studies say be to bias the risk estisates downward. While potential confounding can never be ruled out, a thorough review of the epideaiology studies suggests that it is highly unlikely that the observed increases in CVD risk from ETS exposure are due to confounding. The aajority of studies controlled for at least the major CVD risk factors, and there is a strong consistency of results across different studies from different countries with different lifestyles and diets. Furthermore, a number of studies reported positive dose- response relationships, which are difficult to explain in terms of confounding. 7. Even if the Draft's review were an accurate representation of the available data .and even if its superficial and hypothetical concerns about the reported risks were more convincing, the Draft's conclusion that "the ETS heart disease wholl risk should probably be viewed with soae skepticisa" isl inadequate in view of the potential magnitude of the healt~ risk. k sore appropriate conclusion would be "Because of the known causal association between activ. smoking and CVD, observed even at the lowest doses, the increased risks of CVD from STS exposure consistently observed in epidemiology studies, and the mechanistic data demonstrating cardiovascular effects from low exposures to tobacco smoke,- and because of the potential for substantial morbidity and aortality, the ETS heart disease risk should be viewed with eztreae concernl" I I I I I I I I I I I I I I I