Tobacco Documents Online

A Brief Sunsmary of the Public Commeats Submitted to the US. EPA in Response to the Eaesnil Review+ Draft EPA/600/6-90/0d6A HFXLTH EFFECTS OF PASSIVE SMOIQNG: LUNG CANCER IN ADULTS AND RESPIRATORY DISORDrRS IN CHILDREN Science Advisory Board Meeting December 4 and 5,1990

A Brief Summary of the Public Comments Submitted to the U.S. EPA in Response to the External Review Draft EPA/600/6-90/006A HEALTH EFfECTS OF PASSIVE SMOIGriG: LUNG CANCER nd ADULTS AND RESPZRATORY DISORDERS IN C'HII.DREN L PREFACE Tbe portion of the EPA draft report (the Report) on hing cancer eonsists of I) hazard identification (Is ETS a lung carcinogen? What EPA weight-of-evidence classification should be assigaed, ie, Group A. B, C, D or E?) and 2) quantitative assessment of population rislt (What is the magnitude of risk to the US. popuiation, as measured by exacss lung cancer deaths due to passnn+e smokin& if ETS has been identified as a lung carcinogen?) With regard to hazard identification, the Report first concludes that there is human evidence of an assoaation between ETS exposure and lung cancer occursence. Coupled with knowledge that sidestream smoke contains several known and suspected carcinogens found in mainstream smoke and that tbere is substaatial6uman evidence to support a causal relationship between active smoking and lung cancer, the Report further concludes that exposure to ETS inaeases lung cancer risk and classifies ETS as a Group A carcinogen. Tbe Group A classification (known human carcinogen) refers only to the type of evidence (sufficient evidence from epidemiolo®c studies to support a causal association); no conclusions regarding potency, exposure concentrations, or number of persons at risk are required or implied. 7bese elements are all components of population risk. As in most risk assessments, attempts to quantitatively cbaracterize risk to a population introduce additional sauroe: of uncertainty that, aside from statistical variability, are generally oot very amenable to numerical expression. Although estimation of population risk is usually not attempted for eompounds ctsssified lo.ver than Group A or B m the har.ard identification pfiase, the two prooedures are essentially distinct. Consequently, a revie.vcr's comment will often be more relevant to oae of tbese topics or the other. 2

Unlike evaluation and dassi6ation of potential carcinogens, there is no,direct.veight-of-cvidence dassificatioo scheme for the noo-caneer health related endpoints from the studies of children. The objeaivc of the Report on respiratory disorders in chBdrea ta~itaed to household smohing, particularly lnfants whose mothers smoke near them, is to review, summarize, and assess the cpidcmiologic evidence. The total ewidence of an assadation with ETS e:pocure is persuasive for some endpoints, but the biological basis to anpport a eondusion of causality is weak. Additionally, latent eQeets from smohin,g during pregnancy is a oodaimding factor in almost all sindies. 2. DNIRODUCTION Foltowisg release of the Report on June 25,1990, the EPA received and fdkd requests for over 3600 copies of tbe Report. Over 200 comments from the public.vere submitted by the end of tbe public eomment period, Oct.1,199Q with the bulk of these pertaining to the chapters on lung cancer in adults. Although relatively fewer respondents addressed the chapter on respiratory disorders in children, some kngthy and detailed sets of comments were received on it as well. A number of reviews basically support the Report's conclusions on lung cancer, generally with some constructive criticism. A larger number of reviews, however, are quite kngaby, detailed, and highly aitical. A few reviewers feel' that some aoodusions on respiratory disorders should be strengtbened, with at least two persons referring specifically to asthma. Like the comments on lung caDcer, however, the preponderance of these are highly negative toward the Report. A brief overview is presented below followed by a breakdown by specific topics. The predominant theme in oxmments related to lung cancer is that the cLssifiatioo of ETS as a Group A carcinogen, causally related to inareasad risk of hurg cancer, is unwarranted (Group D elaaifiution it sometimes explicitly recommended instead). Tbe epidemiologic studies are heterogeneous and thus gosuitable for a statistical analysis that oombines their results. If one did so anyway, only the U. S. studies . should be used and the unpublished study by Varela should be included. The resultant overall relative risk is not ugriifiant. Issues of confounding and bias should be addressed to a much greater degree and by individual study: Reviewers disagree on wfiether there is evidence of a dose response ia the studies as a 3

whole. The model to adjust an overaD, risk for smoker misdassif,cstion bias can be improved upon. The Report uses parameter values for the U. S. population in its misclassification model that would yield an atimated overall relative risk of L14 when the true value is L0, i.e., when the overall true accss relative risk due to ETS espocure is zero. Alternative estimates submitted by revie~ ers include L00, L19, and the raage 1?S-L30. The upward adjustment for risk from background eytpocnre to ETS is not needed (because ETS is •ot a lung unoer hazard) and is not c+edibte (beeause of the tse of aotiaioe as a turrogate for ETS and the a::umptioa of liaear dose response at background Iere4 of E'IS, i.e,, due to sources other than spousal amokdag). The claim that it is biotogieaDy plausible that EIS is a lung araiaogen is oenteatious and unjustified in the Report. Regarding the aection of the Report on respiratory disorders in ehOdren, reviewers felt that additional studies should be included, and that eac3 study should be reviewed more carefully for quality, potential sources of confounding and bias, and endibility of study cooclusions. It is not posuble to summariu comprehensively in just a few pages the remarks from the thousands of pages of oommenta submitted. We have attempted, however, to objectively represent the major positions and their associated arguments, particularly those most at odds with the conclusions of,tbe Report. It is felt that this perspective will serve the review proeess most effectively and contribute to constructive eevision as warranted. Tbe material is largely preseated from the perspective of the reviewer,.vithout elaboration e=pt for iosertioa of occasional for clarification. Reviewers' remarks are sometimes lifted directly, or paraphrased, for adoption into this summary when they appear to be a good representation of the comments central to an issue. (fhis is largely an aid to fadlitate eoestruction of this summary and to accurately portray substance.) The discussion on a topic often contains conflicting comments of reviewers, so continuity should not be exputed. Citations are intentionally omitted; it would aot be posswble to cite an the revie+Ners who made similar comments on many topks. C&mrnentarv hv the author of this summarv has been gut in italics in placaes to seDarate it rlearlv from the Meseatatioa of revie.ver's eomments. Tbe major headinp and their order is as foUows: Impropriety of EPw; the .veight-of-tvidence supporting the conclusion for lung cancer, topics omitted or addressed inadequately 4

(lung eancer); topics treated inaccurately or inappropriately Oung canacr); issues related to respiratory disorders in children. 3. AOROpRMTY OF EPA nhe p.aoiraaon of thW tea Pa,t hoc Jo.wWd Lr pnsotaa /kxn the aranapoisr of the revie*w, w*h oonunauary far elonfcatiat rnraspastd be 1sa1'us. 'ltk EPA first began preparation of its workptace smoking guidelines and then eontraated work for a risk a:sessment to support the Aiency: preoonoeivod oondusion that ETS is a lung carcinogen. To support that eoadusion, the aooompanyiqY health risk assessmeat, i.e., the Report, selectively omitted studies (particularly the Varela study), ignored the IIaws of the studies aulixed, oied iaappropriate methods of analysis (particularly meta-analysis) and drew utuuppcxted coodusions by applyiog data for oon-US. studies to the US. population. Additionally, EPA was arbitrary and capricious in fat'bbg to follow its own guidelines for carcinogen risk assessment ,.vhicb requires consideration of all sources of information in evaluating the weight of evidence for bazard identif ation (physical- chemical propertiet, toiocolopc effects, short-term tests, long-term animal uudies, and human uudies.) EPA has not followed the same risk assessment procedure apparent in other chemicals classified as tarcioogens. No long-term controlled animal studies were conducted to provide data at known concentrations of ETS for evaluation of carcinogenicity and dose-response assessment. Uncharuxeriuicaily, the Report relies almost wholly oo human studies, with only a surrogate measure of ETS esposure. Even then, EPA does not follow its own 1989 workshop guidelines for drawing conclusions from epidcmiolopc data. M wvrksJiop doa not consaaete ETA-b,nred pdderina. Tbe Report does M satisfy EYA's three erueria for inferring a causal association Gom human studies, ie, adequately considering and diminatimg bias, eonfound'ung, and the likelihood of ehance as atptaaatioaa. The Report does not follow EPA guidelines for characterizing the weight of evideooe, ie, characxerizing the evidence from human studies and from animal studies individually, combining these characterizations into an overall we& of evidence for human eardnogenicity, and then evaluating the supporting evidence available to determine if the averaU weight of evidence should be modified. 5

EPA exceedcd its statutory authority from Tu1e IV of SIIperfund to tvvArc6 indoor air quality issues (but aot to regulate) and did not act in accordance with law. 7Le Report is an abuse of the Ageacy: discretion. As a health risk assessment, the report mifleads the public by m5dag policy prcfereaces with analysis and oMer•intcrpretiag the data. 4. THE WEIGHT OF EVIDENCE SUPPORTING THE OONCIUSION FOR LUNG GNCBR The topics in the }n!lowing sections conr.tpand to die Jfrst Jot+r statemaitt in du Repoat (p. 1-3 and J- l) descrsbiq the wrig/u of tVidrncr aupportirrg the daui/icatear of E75 as a G+vrp A carcinoren. lIu nanQining gwo statenunts on wright of evidencr on (1) Brond-ba:ed evidence and (2) EA'rcu rtmaiR ofur &djusoncnt for pottruidbics. Comments ntlaud to (1) pnn&*pcr;ain to hctaagentiry of Mdies (sss Section S.l~ sources of bias and confounding (see Secriau S.2 And S.3), and the claimed con.sisaacy of mpo+su (see Section 4-2). Remarks refated so (2) coocan the application of nuca-o,ralysit (see Section 6.2) and posenacl bias from mitnpovred srno+Ying status (see Section 6.3). 4.1. BIOLOGICAL PIAUSIBIIJTY The physical, chemical, and biological propertia of ETS have not been .veD characterized under real-life eavuonmental conditions,.vbere the concern regarding a potential bealtb hazard liec It is inaccurate to assume that aidestteam smoke is, for the most part, the same as ETS. The report notes 43 ideatifud ardaogens in tobaooo smoke tbat are in tidestream smoke, without dariTscatiog that these are not wxzurily related to devdopment of lung aecer in bnmaas and that some bear ao irelevaace to the Report's :ubjer,t. Only about 50 substances have been consistently identiGed in ETS noder real-Iife conditions (compared to bumdreds for sidestream smoke) and the alleged carcinogens have not been found . with consistency or certainty in doses that wrould reasonably be considered to pose a risk to Lumans. Active and passive smoking are vastly dissimilar with regard to exposure concentrations as well. Differences in breathing patterns and duration of exposure, as well as coocentratioas. likely affect uptake and F0 6

dcposition. Valid biolo®cal markers, including the use of DNA or protein adducxs, are not available for comparing active and passive smoking. Furthermore, adequate measures of ETS exposure arc lacking. By eontrast, wme persons auggest a dosimetric approach (called 'cigarcttetquiwaknts' in the Report) to estimate lung cancer risk from ETS expocnre from data oa active smoking An average M apocure is determined to be equivalent to acsiMely smoking some pe:oratage of one cigarette per day. Fssapolating downward on a doae-response auve for active tmoiung at that level at$gests a''oe"bk' lung canxr ri:tc. Zbe Report should aote that it is smokers wW receive the bulk of largely tmdiJuted, fresh sidestrcam smoke ftrom cigarettes and thus are the moct heavily c~oecd passive :mokers. U tidesueam smoke contributes to lung cancer risk in nonsmokers, then it is probably a contributing component in the lung ~ cancer risk associated with active smoking. Three potent carcinogens found in sidestream smoke include benxo(a)pyrene, N-Nitrosodimetbylamine, and N-Niuosodietbylamine. If these substances are carcinogenic at arbitrarily low doses, i.e, if there is no knowrm risk-free exposure level, then it is reasonable to wndude that they pose a hazard to smokers and to ETS- eaposed nonsmokers alilte. •2 CONSISTENCY OF RESPONSE 7his conclusiori has been hearily aftirized at irmccww,c and miskading. The studies are inconsistent in almost every characteristic one could compare (see Section 5.1) and the outcomes differ considerably. Although a aubstantial majority of the available epidemiologic studies have a relative risk estimate exceeding one, tbis comparison alone ignores eftect sizes and euimation unanruinty. T>se preseneo of systematic upward bim e.f., amoker misc]assification (see Section 63), may explain the apparent consistency. The two m4or prospective studies (Hirayama and Garfinkel) are quite diuimihr in most respeus, including their ooac]usions (see Section 6.1). Several revie.ven have suggested that the jU O ` N paragraph on eoasistency of response (Report, p. 1-4) should be retracted. W Q N ~ N ~ W I

<3 UPWARD TREND W DOSE RESPONSE Ahhougjt se+wal nviewers note the Wifuance of an vpward masd to suppon a conclusion of causariry, dtis ropic tat noceiNrd nlasivr[y lrss earenriort in At nHewt, in ratcnal, " some othM e, j, arse of rnaa. eaaysis. Among At nton daailed nriews, At rrina*r ind condusioru xW&V rhe prrssnce of a ornQ mu faconsirre,ic Mun nrvica+ars /Faw adrrcd dut[ At eridatce of an upwdrd ornd is wLUkely so be iu to drmtce. •n auociction (but not eausad auoriarioa) ir conchrded Only seven of the thirteen attdua which attempted to ditinguisb eatpoarre Ieve3s reported a consistent upward trend. One needs to consider whether sources of bias, e.g., due to amoker misdassification, may be dose related and thus eaplaim a false upward trend. The control groups should be omitted in assetsing dose-response trend since they may be qualitatively different in some respects. When that is done, none of the studies exhibit a aigaificant upward trend. Anodur nvieMrw surrnised tJtar evrn wich *e conaaa,oups n,R,ove4 the ovarall Nide,ice is caruister,r with an upper b=cC The Report is incorrect in concluding that there is an upward trend in dose response. After applying two different statistical metbods, one a test of homogeneity and one a test of liaear trend, there is so study.vith a statistically significant trend (at the 5% level) using either method. Given the number of analyses performed, this is quite consistent with there being absolutely no etiect. The Report's conclusions is based on a small number of inadequately reported and analyzed results in the literature, together with a misleading graphical representation of point estimates without consideration of the confidence intervals needed to assess bow statistically different the responses actually are at each of the dose levels. A quite &fferw caulusion is nacbed by miother indepcnQeau analysic included in the reviaws. .T6ere are little data om males but what are avalable show ao evidenee of a trend in ridc. Of fourteen studies evaluated for tread in females, using number of cigarettes smoked per day by the apouse as the esposure surrogate and induding the control dose in the statistical test, seven studies show a significant (p<O.OS) positive dose-related trend, three atudies show some indication of a positive trend, and the remaining four ?0 studies show no apparent relationship. An analysis for trend with duration of exposure from the data N avallable in seven studies shows only two significant outcomes and there is reason to suspect the valid'ity, of~ N ~ 8 N ~ ~

one of them. Overall, the evidence is consistent with somc inaease in risk with amount smoked. Also, risk in the highest exposure goup is higher than that in the lowest exposure poup in 10 out of 14 of the studies and is lower in the remaining four. No ianpGcation of causality s,fiould be d+awn hm. TTu Miewrr finds an msociaiion bn du tprdnniodogk dato but concludes oMadl that the epidcmiologic data do not proyidc omn*sciq tridcnce duN dke a€sociaoon is cmctatAy nlaud so FTS aWwm. •.4. DETECfABLE ASSOCIATION AT ENVIRONMF.NTAL DOSE LEVELS The environmental exposure levels in the studies apply only to exposure of never-smoking women married to tmokers, and these differ between siudies.uith no clear and consistent de5aitan of esposnre to ETS. Exposure to ETS outside the home, including the workplace, is not addressed except in the background adjustment for overaD relative risk. 'lUt adjustment, 6owever, is based on a constructed model tbu inappropriately uses cotiaine biosssays (similar to the cigarette equivalents approach for exorapolating hmg cancer risk from ETS exposure from the dose-response curve for active smoking, with whicb the Report finds fault). The credibility of the Report's a««ment of lnag cancer risk based on epidemiologic evidence ir undermined by the much loa+er estimates obtained from the dosimetric (cigarette eqrzivalents) approach. The Report needs to explain the wide discrepancy in outcomes between these two approaches. Some +m+icwers I+awe faulted the lack of animal data fa dosc-nsponse assessmauf when aposure lnrls to FTS would be bnowrt. Although the Report's assessment of lung cancer risk is based on data for women married to smokers, the population risk obtained for the U. S. indudes never-smoking women not married to a smoker, nev r-smoking men, and former smokers. This exarapolation of figures to an enended population is not Nuified in the repoR, and may be contributing to a large overstatement of populstioo risk. Some nviewrn Aavr felt tlw data for r»alea should lreue been combined with Aar for Jemales, or mnalyud in parnlltF with /ennalc4 in the Rcpoet Scvr>al pasau lurw ogr+ud with lrccpina tlu data for malu and Jemarlu scpmiarr. N . O W .~ rP 9 ~ ~ ~

5. TOPICS OMl'TTED OR INADEQUATELY ADDRESSED (LUNG CANCER) S.L STUDY MTEROGENE]TY Epidemiol'ogic studies differ in a variety of ways, some of which are under the esperimenter's control and some of which ara aot. Those under an experimeater's control iadude all aspects of design, etoectttion, tollo+v-up, and analysis. fsseatially aD the characteristia that determine stndy quality falW into this eategory. L~aaontrotlabk study differeaoes result firom .ariable eharacteristia of the populatioa: to which studies apply. For esample, studies use sponsal smoking as a ansogate to differeatiate between ETS exposure. The studies differ, bo+MCver, in the question(s) asked regarding exposure to spousal smokiag, the method of iatervie.v, and the acceptance of proxy responses. These variables are essentially under the experimenter's control. The true (but unknown risk olluag caacer from aoa•ETS sources, and the true relative risk of ETS exposure, may differ between studies due to culture, race,, ethaicity, lifestyle, diet, presence of other pollutants, and exposure to ETS in the workplace and elsewhere outside the home, which are largely beyond the esperimeater's control. All sources of study heterogeneity are problematic in an overall asseument from the available epidemiologic data. The Report should consider the quality of each study indiydually, including an assessment of potential sources of bias and confounding factors (see Sections 52 and S3). Adjustments for bias, e.g.,, smoker misclassification, should be made at the outset to mdevidual studies, rather than after some aggregation, as from meta-aaalysit (s4e Sections 62 and 63). S2 CONFOUNDING Causes of hrog aaar other than ETS may be responsible for a significant association asoibed to ETS, if not takea into account in the design and analysis of a study. Even a study that takes all potetttial oonfounders thought of into effect may inadvertently be excluding the unknown causal factor. Some potential confounders may result from use of spousal smoking instead of actual ETS exposure, e.g., marriage/divorce patterns, size of homes, proximity to smokers, climate and ventilation. The potential influence of factors 10