Tobacco Documents Online

REVIc, W OF THE U.S. ENVIRONMENTAL PROTEC- TION AGENCY'S TOBACCO AND SMOKE STUDY HEARING BEFORE THE SUBCOiVMITTEE ON SPECIALTY CROPS AND NATURAL RESOURCES OF THE COMMITTEE ON AGRICULTURE HOUSE OF REPRESENTATNES ONE HUNDRED THIRD CONGRESS FIRST SESSION JULY 21, 1993 Serial No. 103-26 Printed for the use of the Committee on Agriculture U.S. GOVERNMENT PRINTING OFFICE 72-758 WASHINGTON : 1993 For sale by the U.S. Government Printing Office Superintendent of Documents, Congressional Sales Office, Washington, DC 20402 ISBN 0-16-041698-1

6 Mr. ROSE. Our first panel is Dr. William H. Farland, Director of the Office of Health and Environmental Assessment, Office of Re- search and Development, Environmental Protection Agency. He is accompanied by Dr. Steven Bayard, the Project Manager for ETS Risk Assessment, Office of Health and Environmental As- sessment, Office of Research and Development, Environmental Pro- tection Agency, Washington, DC; Dr. Hugh McKinnon, Director of the Human Health Assessment Group, Office of Health and Envi- ronmental Assessment, Office of Research and Development, Envi- ronmental Protection Agency, Washington DC. I want to thank you for being here. There were others from your Agency that we had hoped would be with us. Mr. Waxman changed the date of his hearing from to- morrow until today and was able to take several of our witnesses away from us. But we will do the best we can in no way diminishing your abili- ties and your statements. We thank you for being here. Dr. Farland. STATEMENT OF WILLIAM H. FARLAND, DIRECTOR, OFFICE OF HEALTH AND ENVIRONMENTAL ASSESSMENT, OFFICE OF RESEARCH AND DEVELOPMENT, U.S. ENVIRONMENTAL PRO- TECTION AGENCY, ACCOMPANIED BY STEVEN BAYARD, PROJECT MANAGER FOR ETS RISK ASSESSMENT, AND HUGH W. McKINNON, M.D., DIRECTOR, HUMAN HEALTH ASSESS- MENT GROUP Mr. FARLAND. Good morning, Mr. Chairman and members of the subcommittee. Thank you for the opportunity to appear before you today to discuss scientific and procedural issues regarding EPA's report on passive smoking. As you mentioned, I am accompanied today by Dr. Steven Bay- ard a biostatistician in our Human Health Assessment Group, who is the Project Manager, and one of the primary authors of the re- port. I also have with me Dr. Hugh McKinnon, a public health physi- cian who is Director of our Human Health Assessment Group. As you are aware, the U.S. Environmental Protection Agency published an assessment of the respiratory health risks of passive amoking in January of this year. The document has been prepared under the authority granted to the EPA Administrator, including title IV of the Superfund Amendments and Reauthorization Act of 1986-Radon Gas and Indoor Air Quality Research-which directs EPA to conduct research and disseminate information on all as- pects of indoor air quality. The report which was reviewed extensively by scientists from outside of the EPA concludes that exposure to environmental to- bacco smoke or ETS, commonly known as secondhand smoke, is re- sponsible for approximately 3,000 lung cancer deaths each year in nonsmoking adults in the United States and seriously affects the respiratory health of hundreds of thousands of children. My written testimony summarizes the development of the report, the scientific review process, the major findings, and the scientific approach. The 7 testimony concludes with some responses to several tobacco indus- try criticisms of the report. In recent years, comparative risk studies performed by EPA and its Science Advisory Board have consistently ranked indoor air pol- lution among the top five environmental risks to public health. As part of its efforts to address all types of indoor air pollution, EPA's Indoor Air Division in 1988 requested the EPA's Office of Research and Development to undertake an assessment of the respiratory health effects of passive smoking. Because of both resource and time limitations, the assessment was limited to respiratory health effects, both cancer and noncancer. The report was prepared by my office, the Office of Health and Environmental Assessment within the Office of Re- search and Development, and was written by both in-house staff and outside contracting assistance. Before being released in draft form for public review, the passive smoking report received many internal reviews, mostly from within the Office of Research and Development. Various parts of it were also reviewed by selected outside experts, both from other Federal agencies and from academic institutions. Revisions incorporated the reviewers' comments wherever possible. The first external draft of this assessment was released for pub- lic review and comment in June of 1990. In December 1990, EPA's Science Advisory Board, a committee of independent outside sci- entists, conducted a review of the draft report and submitted its comments to the EPA Administrator in April of 1991. In its comments, the SAB's Indoor Air Quality/Total Human Ex- posure Committee concurred with the primary findings of the re- port, but also made a number of recommendations for strengthen- ing it. Incorporating recommendations from both the public and the Science Advisory Board, a revised draft was transmitted to the board in May of 1992 for a second review. Following a July 1992 meeting, the SAB panel endorsed the report and its conclusions, in- cluding a unanimous endorsement of the classification of environ- mental tobacco smoke as a group A or known human carcinogen. EPA also received and reviewed ~public comments on the second draft and integrated all appropriate material into the final risk as- sessment. The final report was released in January of this year at a joint press conference held by former Administrator Reilly and former Department of Health and Human Services Secretary Sulli- van. Based on the weight of the available scientific evidence, EPA has concluded that widespread exposure to environmental tobacco smoke in the United States presents a serious and substantial pub- lic health risk. In adults, ETS is a human lung carcinogen, responsible for ap- proximately 3,000 lung cancer deaths annually in U.S: nonsmokers. ETS has been classified as a known human carcinogen under EPA's carcinogen assessment guidelines. This classification is re- served for those compounds or mixtures which have the strongest data to determine a cause-and-effect relationship, including data from human populations. Only 10 to 15 other agents, including as- bestos and radon, have been classified by EPA as group A carcino- YrT8MV09

8 gens, and ETS is the only one for which cancer has been observed at typ ical nonoccupational environmental levels. ETS has subtle but significant effects on the respiratory health of nonsmokers including coughing, phlegm production, chest dis- comfort, and reduced lung function. In children, ETS exposure increases the risk of lower respiratory tract infections such as bronchitis and pneumonia. Our estimates are that between 150,000 and 300,000 of these cases annually in infants and young children up to 18 months of age are attributable to exposure to ETS. Of these, between 7,500 and 15,000 are esti- mated to result in hospitalization. In addition, ETS exposure increases the prevalence of fluid in the middle ear, a sign of chronic middle ear disease. Fluid in the mid- dle ear is a major cause of hospitalization- of young children for an operation in the United States. ETS exposure in children irritates the upper respiratory tract and is associated with a small but significant reduction in lung function. In addition, ETS exposure increases the frequency of episodes and severity of symptoms in asthmatic children. The report esti- mates that 200 000 to 1 million asthmatic children have their con- dition worsenedby exposure to environmental tobacco smoke• and ETS exposure is a risk factor for new cases of asthma in chiidren who have not previously displayed symptoms. EPA reached its conclusions concerning the potential for ETS to act as a human carcinogen based on an analysis of all available data. Specifically, the finding that EPA should be classified as a group A carcinogen is based on the conclusive evidence of the dose- related lung carcinogenicity of mainstream smoke in active smok- ers, the chemical similarities of mainstream smoke and the side stream smoke given off the burning end of the cigarette and the known exposure and uptake of ETS at levels which could increase risk. The finding is bolstered by the statistically significant exposure- related increase in lung cancer in nonsmoking spouses of smokers which is observed in analysis of more than 30 epidemiology studies from eight different countries that examined the association be- tween secondhand smoke and lung cancer. The weight of the evi- dence analysis for the noncancer respiratory effects in children is based primarily on a review of more than 100 studies, including over 50 recent epidemiology studies of children whose parents smoke. EPA is not the only Federal agency that has evaluated environ- mental smoke. The EPA's conclusions on the respiratory effects of passive smokix~~g strengthen and confirm those of earlier assess- ments by the U.S. Surgeon General in 1986 and the National Re- search Council of the National Academy of Sciences in 1986. The World Health Organization has also concluded that ETS causes ex- cess risk of lung cancer in 1986 and other respiratory disorders in 1992. The National Institute of Occupational Safety and Health in 1991 concluded that occupational exposure to ETS causes increased risk of lung cancer and probable heart disease. The position of the National Cancer Institute in 1993 is that ETS is a proven cause of lung cancer in nonsmoking adults and is asso- 7VZM9v0g 9 ciated with an increased risk of coronary heart disease. Since the cutoff date for literature inclusion in the EPA report, several new studies have been published which provide additional evidence of respiratory effects from ETS exposure. Six of these are particularly relevant, one each on sudden infant death syndrome, SIDS, and asthma, and four on lung cancer. Three of the recent studies on ETS exposure and lung cancer in nonsmoking women add to the data base of the 30 studies analyzed in the EPA report. Two of these, Stockwell, et al. from the Journal of the National Cancer Institute and Brownson, et al. from the American Journal of Public Health are large U.S. case-control stud- ies which find significant increased risks among nonsmoking women in the highest category of ETS exposure based on the amount their husbands smoke. Similar results are reported in the very recent study of non- smoking Chinese women by Liu, et al., in the American Journal of Epidemiology; Liu also found a statistically significant increase in risk in the most exposed group, based on husband's smoking. In ad- dition, Stockwell, et al. found significantly increased risks for high levels of household exposure in children. I believe it is important that we put these risks associated with ETS in perspective. The EPA estimates that about 20 to 30 percent of all lung cancers caused by factors other than smoking are attrib- utable to environmental tobacco smoke. Another way of expressing this is that the increased risk of dying from lung cancer is about 1 in 1,000 from all ETS exposures out- side the home. For reference, a one-pack-a-day smoker experiences lung cancer risks approximately 100 times higher or a 1-in-10 risk. Exposure to ETS varies, but higher exposures are associated with higher risk. For example, people whose spouses smoke in the home face an average increased risk of 2 in 1,000. Estimated risks in this range are considered high. For comparison, EPA generally sets its stand- ards or regulations so that increased cancer risks are below 1 in 10,000 to 1 in 1 million. , In other words, the increased lung cancer risks associated with exposure to environmental tobacco smoke are at least an order of magnitude greater than the cancer risks for virtually any other chemical or agent that EPA regulates. The additional risks on childhood respiratory health make an even inore compelling case for the public health impact of ETS. In my written testimony I have also addressed many of the criticisms of EPA's approach to and findings of this analysis. We will now be pleased to answer questions from the subcommit- tee. [The prepared statement of Mr. Farland appears at the conclu- sion of the hearing.] Mr. RosE. Thank you very much, Dr. Farland. Could you describe to us EPA's guidelines for carcinogen risk as- sessment, and in particular, what those guidelines say with regard to the classification of a substance as a group A carcinogen? Mr. FARLArro. Yes, Mr. Chairman.

10 The Agency has a long history of documentation of its guidance on carcinogen assessment. It put out its first interim guidance in 1976 and followed with final guidance that we use today in 1986. That particular guidance is used by scientists within the Agency, and by those outside to understand how the information is going to be translated into risk assessments, and to help us with the work that we do in my office. We use a weight of evidence approach. This particular set of guidance stresses the importance of considering all the information, and it uses a classification scheme as one aspect of the guidance that breaks carcinogens down according to the understanding of the evidence available on them. They are categorized as A, known human carcinogens; B, probable human carcmogens; C, possibly human carcinogens; and then there is a D category, which is not classifiable; and an E category for those chemicals that have been sufficiently well tested and don't show a carcinogen response. The A carcinogen class is our highest class of evidence. It gen- erally will include human information, epidemiology studies, infor- mation on human metabolism, and other types of human data. It will have looked carefully at those epidemiology studies in order to attempt to rule out confounders that might cloud the analysis of those particular studies. This is a classification system that has been broken up into these five categories. Mr. ROSE. I understand that. You have been over the classifica- tion system. Is it not true that in your guidelines that you indicate that for a substance to be classified as a group A carcinogen there must be sufficient data in humans, that is, epidemiological data? Mr. Fnttr..ANn. The type of information that we are talking about is all of the human data. We are looking for a classification that would include sufficient human data. Mr. ROSE. You contend that you have sufficient human data? Mr. FARLAND. We do. That sufficient human data includes infor- mation on active smoking in humans, it includes information on the eXposure of humans to- Mr. ROSE. What are some other class A carcinogens? Mr. FAximn. The Ag ency has classified between 10 and 15, radon, benzidine, byschloromethyl ether, vinyl chloride, asbestos, nickel, arsenic-these are chemicals that have reached the cat- egorization of being known human carcinogens. Mr. ROSE. What are some class B carcinogens? Mr. FARLAND. Class B carcinogens have an adequate animal data base. Mr. ROSE. What are they? Mr. FARLAND. Formaldehyde- Mr. ROSE. So tobacco smoke is more carcinogenic than formalde- hyde? Mr. FARLMn. It is not a question of more carcinogenic; it is the data base that is available. Mr. ROSE. Environmental tobacco smoke is A and what was the classification for B? Mr. FARLAND. That is a probable human carcinogen. Formalde- hyde falls into that category and there is limited human e~idPncw on formaldehyde. IGUTMM9 11 Mr. ROSE. It might be poisonous if you drank it, but you are talk- ing about the vapor from it; is that correct? Mr. FnRLANn. Yes. There is limited human evidence that it causes upper respiratory tumors. Mr. ROSE. How was a determination as to formaldehyde arrived at? Was it done through a comparison of various studies that were done around the world or was there a particular test that you all conducted? Mr. FAxt.AND. The formaldehyde conclusion- Mr. ROSE. You were probably looking at studies that had been done in various places? Mr. FARLAND. That is right. Mr. RosE. Did you apply the 95 percent or the 90 percent con- fidence level in the case of formaldehyde? Mr. FARLAND. I would have to go back to look at that. The test for statistical significance in those cases would have been 95 per- cent. Mr. RosE. You applied 90 percent to environmental tobacco smoke which made it a class A carcinogen, but you haven't looked at formaldehyde or other substances at the 90 percent level? Mr. FARLArrD. No. The test for statistical significance is at the 95 percent level in both cases. Mr. ROSE. What did you change to 90 percent? I am not obvi- ously asking the right questions, Doctor. What did you change from 95 to 90 percent? Mr. FARLAND. The important issue is that the statistical significance- Mr. ROSE. What was the level that you changed from 95 to 90 percent in the case of environmental tobacco smoke? Mr. FARLAND. We used a 95 percent statistical significance test that is one-tailed and resulted in a 90 percent confidence interval. It is a standard statistical procedure. It is the confidence interval that comes from the 95 percent statistical test and is one-tailed. Mr. ROSE. So you did not use a 90 percent confidence downgrade in your conclusions as to environmental tobacco smoke? Mr. FARLAND. This is not a downgrade. A 90 percent confidence interval is not a downgrade. A 90 percent confidence interval is consistent with a 95 percent one-tailed test. And that is what we used, a 95 percent one-tailed test of significance to determine whether or not the observed relative risk was significant compared with controls. Mr. ROSE. All right. I better go back and learn new math. I thought 95 was higher than 90, but I am wrong. Mr. FARLAND. Mr. Chairman, you are correct that a 95 percent confidence interval is a more stringent statistical test and there are 95 percent statistical significance tests that use a 95 percent con- fidence interval. But in the case where one has evidence that the effect is likely to be adverse and not beneficial, the statistical use of a one-tailed test is appropriate. Mr ROSE. So you assumed it was bad and did a one-tailed test? %Ir FAHt.A*rD _ We assumed based on the evidence of lung cancer Nni mmAing thet the result would likely be an adverse effect and ...- %,..A a onN tailed test of significance.

12. Mr. ROSE. But based on your studies and the way you classified them, you are telling us that the fumes from tobacco smoke are more carcinogenic than the fumes from formaldehydes? Mr. FARLAND. This is a different- Mr. ROSE. There are people out there who obviously work for you who are shaking their heads. Mr. FARLAND. It is not a matter of quantification. It is not a mat- ter of it being more carcinogenic or less carcinogenic; it is the weight of the weight of the evidence that we have available on that substance. Mr. ROSE. You are really getting my attention. Mr. FARLAND. A known human carcinogen may not be as potent a carcinogen, as stronp a carcinogen, as one that we don't have human data op. There is a difference between hazard and potency. Mr. ROSE. What does the risk ratio of one mean? Mr. FARLAND. A risk ratio of one means that there is no increase in relative risk so that this is a ratio of Mr. ROSE. Increase? Mr. FARLAND. No increase in relative risk over a background or a control population. Mr. ROSE. So a risk ratio of less than one means what? Mr. FARLAND. A risk ratio of less than one depending on the con- fidence around the estimate, may mean that it is protective or it may mean that it is equivalent to one. Mr. ROSE. Above one, what does that mean? Mr. FARLAND. Again, it means in this case that there would be an increased relative risk or, depending on the confidence, it may be equivalent to one. Mr. ROSE. Looking at some studies that you have used, associa- tion between risk of lung cancer and childhood exposure to tobacco smoke among nonsmoking women, are the risk ratios above or below one? Mr. BAYARD. Are you talking about childhood exposure? Mr. ROSE. I just said that. Association between risk of lung can- cer and childhood exposure to tobacco smoke among nonsmoking women; the risks are all less than one. Mr. BAYARD. Did you look at Stockwell or Janerich? Mr. RosE. This is the Fontham study. Mr. BAYARD. That study showed no increase in risk in children. The Stockwell study has shown an increase in risk and the Janerich study showed an increase in risk. In the Fontham study, she discusses at the end that her study did not show an increase, and that conflicted with the Janerich study. Mr. ROSE. Then you could conclude that according to the Fontham study, you should expose children to tobacco smoke to re- duce their risk; is that right? Mr. BAYARD. Only if you are willing to conclude by the Fontham study that you should definitely not expose people to ETS at work because they showed a very large increase for people exposed at work. The relative risk for exposure as children was very close to one as I understand it; maybe 0.9-something. Mr. ROSE. So when you use a one-tailed approach as opposed to a two-tailed approach, you don't look at the upside and the down- side; you just assume that there is a danger here? 13 Mr. BAYARD. You assume that if there is any effect-this is for lung cancer-we did it different than for lung cancer than we did in noncancer respiratory effects. For lung cancer, we had evidence that high levels of tobacco smoke caused lung cancer and that is pretty good evidence, about 9 million people studied worldwide; and most people will admit that tobacco smoking causes lung cancer. So we believe that with the lower levels of tobacco smoke from environmental tobacco smoke there would not be a protective effect, but if there were any effect, it would be an adverse effect. So we only used a one-tailed test or 90 percent confidence intervals for the analysis of ETS epidemiology and lung cancer. But for the childhoad respiratory effects, we used 95 percent confidence inter- vals because we didn't have the evidence from smoking in children causing noncancer respiratory effects; so we did do it two ways in the same report. Mr. ROSE. I kind of broke a committee rule here. I should hear the whole panel. I have asked basically 1'rh questions. Are you the only one going tog~ve a statement, Dr. Farland? Mr. FARLAND. Yes, Mr. Chairman. Mr. ROSE. Then I am not wrong. Could you explain what criteria the Agency uses when evaluating epidemiolog~cal studies to deter- mine whether an association could be due to chance? Mr. FAWANn. The Agency uses an approach that was published by Bradford Hill. It is a traditional approach that goes through a number of criteria for causality, and it mcludes a number of things like consistency of the data, the number of studies, the strength of the response the biological plausibility and those sorts of issues. In addition to lat, we use statistical approaches as well as those gen- eral approaches for evaluating causality. Mr. RosE. Would you describe as sufficient any single epidemio- logic study or the combination of a series of such studies that re- ports a relative risk that is statistically not significant? Would you describe as sufficient any single epidemiologic study or the com- bination of a series of such studies that reports a relative risk that is statistically not significant? Mr. FARLAND. Mr. Chairman, the only example that I can think of where we have argued a strong case for known human carcino- genicity where the epidemiology data base may not be statistically significant is vinyl chloride where there is a very limited number of cases, but there are very specific types of cancers so the cause and effect relationship based on the biological arguments is very strong. Mr. RosE. Dr. Farland, at the Science Advisory Board review of ETS risk assessment in July 1992, you stated that ETS risk assess- ment has been an innovative approach and that it bears some merit in terms of future approaches for risk assessment. Does that mean that in the future EPA will place undue emphasis on the much criticized statistical technique of meta-analysis? Does the Agency have guidelines on the use of meta-analysis? And moreover, if the ETS risk assessment is to be a template for future directions, does it mean that the Agency intends to override existing carcinogen risk assessment guidelines in favor of a less strict weight of evidence approach that would leave the Agency much more freedom to interpret data as it might wish? F XVT8MV0Z

14 Mr. FARLAND. I am not sure that I got all those questions. Mr. ROSE. At the Science Advisory Board's review of the risk as- sessment on July 1992, you stated that ETS assessment has been an innovative approach, that it bears some merit in terms of future ap roaches for risk assessment. Sne, does that mean that in the future you will place em hasis on the much-criticized statistical technique of meta-analysis? Two, does the Agency have guidelines on the use of meta-analysis; and three, if environmental tobacco smoke risk assessment is to be a template for future directions, does it mean that the Agency in- tends to override existing carcinogen risk assessment guidelines in favor of a less strict weight of the evidence approach that would leave the Agency much more freedom to interpret data as it might wish? Mr. FAR[.Arrn. Thank you for repeatinP. I made a statement about the innovative approaches that we use. I think this is a unique data base where one has 30 lung cancer studies and over 100 studies in children. Fortunately, we don't have many cases where we have that much human data on an en- vironmental contaminant. With regard to your question on the emphasis on meta-analysis, the use of meta-analysis depends on the quantity and quality of the studies that you have available, whether they can be combined, and there will be few cases where we will have enough information, sufficient data, sufficient numbers of studies of similar design and so on to be able to use meta-analysis. We are committed to use this particular approach which is gaining favor within the epidemiologic and statistical community in future cases where it would fit. In terms of guidelines for meta-analysis, the Agency has none, but it does very carefully lay out the process it used for meta-anal- ysis and subjects it to external peer review as part of its peer re- view process. In addition, we are working with industry and aca- demic groups to hold a workshop on meta-analysis in this upcom- ing year to look at some of the princip:es that would be included in a general guidance document on meta-analysis. In terms of the cancer guidelines, we are in the process of revis- ing our cancergu idelines to move along with the evolution of the scientific data. We will focus on weight of the evidence approaches and will continue to focus on a strong scientific judgment compo- nent within ourgu idance. Mr. RosE. Is there any substance for which EPA has used a set of epidemiologic studies pertaining to a substance other than the one under study, even though perhaps similar in some respects, in order to determine a group A classification? Is there any substance that you have used a set of studies per- taining to a substance other than ETS, even though perhaps simi- lar in some respects, in order to determine a group A classification? The answer is no, isn't it, Doctor? Mr. FARLAND. I think the best example that we have of that is the case of our evaluation of benzidine containing dyes. In that par- ticular case, benzidine is known to be a human carcinogen. Dyes that contain benzidine that are likely to be metabolized, and have been in some cases shown to be metabolized by humans so that benzidine appears in the urine, are considered as known human 15 carcinogens where there is no direct epidemiologic data on those dyes. In that case, we use the surrogate data from benzidine and the epidemiology studies there to make conclusions on these other dyes. Mr. ROSE. The EPA risk assessment on environmental tobacco smoke determined that ETS could be classified as a group A car- cinogen solely on the basis of comparisons with the epidemiology of active smoking. This is at odds_ with the recommendations of EPA's Science Advisory Board. Could you explain why in this instance EPA decided to reject the advice of the Science Advisory Board? Mr. FAxt.ArID. Let me comment and perhaps Dr. Bayard would like to add an additional comment. My understanding was that the Science Advisory Board suggested to us that although we had spent a lot of time evaluating the epidemiology studies, that it was pos- sible to make a conclusion simply on the similarities between active and passive smoking. They didn't suggest that we should do that instead of evaluating the epidemiology studies; they just said that the strongest case would be based on both of those and that we should go back and improve the arguments on the correlation with active smoking in addition to our analysis of epidemiology studies. Mr. BAYARD. The statement you are referring to I think was made by Dr. Lippman at the second Science Advisory Board meet- ing July 22, 1992. That statement, as a lot of statements made at that meeting, did not get into the Science Advisory Board report to us on November 20, 1992, so that oral statement did not represent the Science Advisory Board's consensus. Our conclusion was that we could label environmental tobacco smoke a known human carcinogen based on the similarity of envi- ronmental tobacco smoke to mainstream smoke, with both contain- ing the same carcinogen, and the known lung cancer response from mainstream smoke down to very low doses with no evidence of a threshold. This conclusion was in the draft which we sent to the SAB for review, and to which that report of November 20 referred. So our conclusions were in the second draft which went to the Science Advisory Board for review. They agreed with the conclu- sions in their November 20 report to us, and it remained in our final report which was published in December. Mr. ROSE. I have a lot more questions for Dr. Farland, and for Dr. Bayard, but in fairness to the members of the panel, I am going to yield to Mr. Lewis for questions and then to my colleagues on my right and then Mr. Lewis, whoever he will wish to recognize. Mr. Lewis. Mr. LEwis. Thank you, Mr. Chairman. Dr. Farland, Dr. Shapiro of the Sloan Epidemiological Unit stat- ed in a paper that the use of meta-analysis and observational re- search should be abandoned and the guidelines also go on to state that negative results from a well-designed and well-conducted study that contains usable exposure data can be used to define the uppEr limits of risk. What implications does this have for a series of studies where several of the larger studies report no increase in risk? UT8~VMZ

16 Mr. FARLAND. As I mentioned there is controversy with regard to the use of meta-analysis, but it is growing in favor in terms of trying to combine studies to increase their power to evaluate ef- fects. There are individuals in the scientific community who are not comfortable with the use of meta-analysis and feel that perhaps it should not be used. There are others who have used it extensively to evaluate car- cinogens. A report by Sir Richard Doll, for instance, used meta- analysis extensively and suggests that it is the best way to evalu- ate a particular class of compounds that he is looking at. The impli- cations of the meta-analysis approach is that one can combine stud- ies showing an effect, that have an increased relative risk, with those that could not show an effect and get some sort of a sense as to the contribution of that no effect finding on the overall esti- mates of relative risk. There is a rationale for doing that sort of thing, and there is a basis for, rather than focusing only on the positive studies, using the positive and nonpositive studies, the no relative risk increase studies, in trying to reach your conclusions. That is what meta- analysis helps us to do. Mr. LEWIS. One of the largest studies by Brownson reports no in- crease in risk. This should have a significant implication for the risk assessment, is that true? Mr. BAYARD. The Brownson study is 1 of 33 studies on lung can- cer and environmental tobacco smoke among never-smoking women. The Brownson study found a significant risk among the women most heavily exposed. If you take all the women, the women who were ever exposed versus those never exposed to their husband's smoke, the Brownson study found no overall increase in risk. You have to understand that these epidemiology studies done at true environmental levels are not the easiest studies to detect any kind of an effect. In fact when EPA calls something a known human carcinogen, most often the studies are based on high occu- pational levels, anywhere from 100 to 1,000 times what a typical environmental level will be. ETS is the only agent which EPA has ever found to be, ever declared, a known human carcinogen, which has actually been found to be carcinogenic at true environmental levels. Getting back to the Brownson study, Brownson concluded that ours and other recent studies suggest a small but consistent in- creased risk of lung cancer from passive smoking. Two other stud- ies which have also appeared since our cutoff date for literature re- view have also found statistically significant increases at the high- est level and one found a significant dose response trend. The question is would Brownson have changed our results? The answer is no. That would have been 1 of 33 epidemiology studies on ETS and lung cancer, but there are tons of other studies that also went into our weight of evidence, hundreds and hundreds of other studies. The answer is no. Mr. LEwIS. Is there any other substance for which EPA has used a set of epidemiology studies pertaining to the substance other than the one under study even though perhaps similar in order to determine a group A classification? M8Mfn i 17 Mr. BAYARD. Nickel causes lung cancer and nasal cancer in pyrometalurgical refinery workers. Those people are exposed to high doses of nickel which also probably contains sulfuric acid, so we have found that, yes, some of these nickel salts are carcmo- genic. But do we know if it is ever going to cause lung cancer at typical environmental levels? We don't know that. We have never seen- with the exception of environmental smoke-cancers from our group A carcinogens at typical environmental levels that we know about. Asbestos, we have never seen cancer from background levels in schools. We know we spend a lot of money to clean them up, but we have really never seen cancer at these levels in schools. We hypothesize that that is going to be the case based on model- ing but that is not true for environmental tobacco smoke. So nickel is one. Coke ovens is one. Coke oven workers who have high expo- sure to coke oven emissions come down with lung cancers. By the time these emissions have dissipated and get into the ambient en- vironment we don't particularly know if they are actually going to cause lung cancer. Does that answer your question? Mr. LEwIs. I think it does because I really don't know what you used. It was my understanding you only used environmental to- bacco smoke. Apparently you are telling the subcommittee that you did use others for companson. Mr. BAYARD. The question you asked was whether there were any other chemicals for which we used like studies, I thou ght that meant, in which the studies that we used were actually different from what was available in the environment. My response was that, yes, with both nickel and coke ovens we had occu~ pational exposures which are probably going to be some- what different from what the environmental exposures would be; not only in dose, but in physical chemical characteristics. Mr. FARLAND. Mr. Lewis, I also mentioned to the chairman the idea that in the case of some of the benzidine containing dyes, we had used epidemiology studies from benzidine in order to classify those specific dyes. That would be another case where we have used surrogate data. Mr. LEwis. After EPA's adjustment to the 90 percent confidence interval, how many of the studies reported statistically significant increase in risk? Mr. FARLAND. Could I use the chart that we had up here? That would be helpful. Mr. LEwis. Fine. Mr. FARLAND. This chart basically just shows the weight of evi- dence approach that we used. There were 30 epidemiologic studies of ETS and lung cancer. If you compared the 30 studies for ever versus never exposed which is the lowest level of evaluation, of the 30 studies, 24 showA an increased relative risk 9 were statistically significant. That is a finding that would not likely be due to chance. The probability of getting 9 statistically significant studies among 30 by chance is a 1 in 10,000 probability. When we broke that group of 30 down into the 17 studies which characterized ex- posure and used that exposure level and looked at the increased

18 risk in the highest exposure level, 17 out of 17 studies showed an increased relative risk, 9 were statistically significant in that group and that is a probable finding of 1 in 10 million. We wanted to see which studies showed a positive dose trend. Of the 14 Studies that showed the characteristics to evaluate dose re- sponse, 10 were statistically significant. By chance, about 1 in 1 billion that you would get 10 out of 14 coming up as a probability of occurrence by chance. Mr. LEwIS. Could I stop you there for a moment and ask you, in the first instance you said nine were statistically significant? Mr. FARLAND. Correct. Mr. LEwts. In the second, 9; and in the third, 10. That was 9 out of 24 or 30, and 9 out of 17 Studies and 10 out of 14. How about the rest? - Mr. FARLAND. The others were studies of lesser power, smaller studies. They showed a nonstatistically significant increase or no increase at all. That could be due to the nature of the study. It could be a true evaluation of that particular test, or it could be due to chance that those results showed no increase. Mr. LEWIS. Wouldn't those studies be significant to bias these nine in some way? Mr. FAIU.AtvD. They continued to raise uncertainty within the overall assessment because not all studies have shown a positive response, but not all studies are equal. They are not designed the same way, looking at the same populatione Mr. LEWIS. You can design a study to be the way you want it. Mr. FARLAND. I would agree that that could be done. I would hope that it would not be done. Mr. LEWIS. At the 95 percent confidence level, Doctor, how many of the United States' ETS epidemiological Studies of spousal emoke exposure to lung cancer report significant results as employed in this risk assessment? You did mention this in your opening statement, but at the 95 percent confidence level, how many? Mr. BAYARD. May I answer that? For the ever versus never e osed, I think there was only one or two, Fontham or Fontham an~ Chorea. Only 1 or 2 out of the 11 showed overall atatistical significance. When you start looking- you would have only expected out of 20, if there is no effect, re- member we talked about the 5 percent significance level. When you deal with a 5 percent significance level, if there is no effect, it would be significant 1 time of 20. If no effect, we would have expected 0.5 of the 11 studies or less than one study to be sta- tistically significant. Of the ever versus never exposed, there were one or two U.S. studies which were statistically significant. When you start looking at trends, two or three were statistically significant and in the highest exposure groups, three were, but a lot of Studies just didn't have the information available to test at the highest exposure groups. So based on the 11 studies, we didn't see a tremendous ef- fect in the United States and we explained that by looking at all different countries and our analysis separated the results by the different countries. 19 Mr. LEWIS. Didn't you find that those studies at the 95 nercent confidence level did not meet the criteria that you are telling me until it was dropped down to 90? Mr. BAYARD. No. That isn't the way I remember it. You are talk- ing about only the 11 U.S. studies? Mr. LEWIS. Yes. Mr. BAYARD. It doesn't matter whether there were two studies, if there were two significant studies at the 95 percent level or one or two with the one-tail test. I don't think there was much of a dif- ference because even at the 95 percent level, there wasn't that much significance; if you did the ever versus never, which is a crude measure. A better measure to take is the highest exposure ou Don't forget, everyone is exposed to ETS. If you take ever versus never, even the people you say are never exposed will be exposed, so you are comparing risks of those with a little bit more exposure to those with a little less exposure. It is hard to define a result from any one study. That is why we tried to take all possible stud- ies, the positive and the negative studies, and tried to see what the story was. Mr. LEWIS. I understand as scientists you have to defend your studies. Mr. BAYARD. They are not my studies. We just did the analysis. Mr. LEWIS. I have a problem as the chairman did, going from 95 to 90, it seems like it was establishing a atatistical modality in order to meet a study requirement. I am not accusing you of that, but it is confusing to me. Mr. BAYAxD. We did it two different ways in the same report. For the childhood respiratory effects, we used a two-tailed test, for the lung cancer analysis we used a one-tailed test based on the prior evidence that active smoking causes lung cancer. So we did it two different ways. Mr. LEWIS. Mr. Chairman, I have one last question for Dr. Bay- ard. You stated at an open meeting in EPA on January 7, 1993, chaired by Mr. Bretthauer, that the use of 90 percent confidence intervals was justified because you had a prior feeling that ETS would cause cancer and so it was appropriate to use a so-called one-tailed test and look only at increased risk. How often has EPA adjusted its statistical standards on the basis of a prior feeling? Mr. BAYARD. We do use 90 percent confidence intervals when we extrapolate from high animal to low human exposure, so in that sense yes, we do use 90 percent confidence intervals. The question of whether we adjusted these intervals to get the desired results, it is just the way we did it. We looked-before we even looked at the data, we said what is our prior belief on environmental tobacco smoke; is it going to be beneficial or adverse? Do we have enough evidence to say it is not going to be beneficial? If you don't know which kind of effect you are going to have, you use a two-tailed test. If you have a strong enough belief that any effect you have is going to be adverse, you use a one-tailed test and that is exactly what we did. My belief is that any effect of environmental tobacco smoke would be an adverse one for lung cancer. 4,VT8~VMZ

20 That is not true for childhood respiratory effects. We didn't have that prior belief. This is standard statistical procedure. It was something we raised to our Science Advisory Board. We did not change the methodology. We used a one-tailed test in the first draft, we used a one-tailed test in the second draft, we used a one- tailed test in the final, for lung cancer. We used a two-tailed test in the first draft for respiratory effects, we used two tailed in the second draft for respiratory effects and we used two tailed in the final. We did not change it. We brought it before the SAB and they examined it and said it is fine. It is standard statistical procedure. You learn it in your first course in statistics. Mr. LEWIS. Mr. Chairman, I didn't want to confuse anything. Mr. ROSE. Mr. Baesler. Mr. BAESLER. From your last statement, Doctor, you did start this study with the feeling that you felt there was a problem? Mr. BAYARD. No. Mr. BAESLER. You just said that. Mr. BAYARD. When I started in 1988 I didn't know anything about- Mr. BAESLER. The previous answer to his question you said that you did a one-tailed test because you had the prior feeling that, first of all, there wasn't anything positive out of tobacco smoke, it was going to be negative, and therefore you did the one-tailed test and you had a prior feeling that there was going to be a problem. Mr. BAYARD. With respect to lung cancer? Mr. BAESLER. Yes. Mr. BAYARD. We felt if there was an effect- Mr. BAESLER. You said you had a prior feeling that the effect would be negative. You didn't say if there was an effect. Mr. BAYARD. If there were an effect of environmental tobacco smoke we did not expect it to be protective- Mr. BAESLER. You are backing up. Nowhere did you put if there was effect. The answer to his question was you had a prior feeling there was going to be a negative effect. That is the way you just answered the question. I think that is a big difference to suggest that if there was an effect. The critics will say that you never went into this, EPA never went into this study with the question if there is an effect. The crit- ics will say you went into the study, there will be an effect, how do we substantiate it? Is that true or not true? When you started the study, yo~u had a prior feeling there would be a negative effect; therefore. Therefore you wanted to use a one-tailed effect-mumbo jumbo, nobody understands, thou gh you have said a great deal, and I don't. Representing the largest burley industry in the country, it concerns me that assumptions that you make so cavalierly you make that can so negatively affect such a large industry, which this feeling that you had when you started this bothers me. You said we had this feeling that there was going to be a nega- tive effect; therefore, we did this one-tailed study, which for what- ever scienctic reason, we went from 90 percent to 95 and nobody understands it, unless it is another scientist. We did that, nothing wrong with that. You answered the question about in 1992 some person ques- tioned on the panel about well, are we doing it right or not and you 21 cavalierly threw that off, that wasn't the consensus of the panel, that was one fellow and we didn't use it until the next two times. We have cavalierly thrown off every dissenting view and that bothers me. I have no idea what you are talking about, haven't un- derstood a thing you said all day other than the fact you are trying to defend a study that basically attacks a large industry and you started with the presumption that it would be negative. That both- ers me, because you are supposed to represent all of us, not just one side. The second concern is you said a minute ago in answer to a ques- tion, something about animals. Am I wrong or right? I have no idea, but do you often in these type of studies, class A, use animal- type testing? You do, don't you? Mr. BAYARD. We look at all the evidence. Mr. BAESLER. Did you in this one? Mr. BAYARD. Yes. We looked at animal evidence. Mr. BAESLER. With respect to smoke? Mr. BAYARD. Yes. Mr. BAESLER. So you did that the same as you do the others; cor- rect? Mr. BAYARD. We looked at animal evidence; yes. Mr. BAESLER. How did it affect the animals? Mr. B4YARD. It is more mutagenic than mainstream smoke when applied to the mouse skin and in cell colony tests. Mr. BAESLER. Maybe it is not possible, but can you just-I think basically maybe I am dumb, but I don't understand a thing you are saying. Maybe that is intentional. Tell us in common terms that somebody will understand what we are talking about. We are not scientists. Mr. ROSE. What does mutagenic mean? Mr. BAYARD. Causes gene changes in the DNA which is thought to be a mechanism related to the start-up of cancer. Mr. BAESLER. We talked about 11 studies that you had used in your approach to this analysis. The meta-analysis, you used 11 studies. Mr. BAYARD. No. We used 30 epidemiology studies in the meta- analysis. Mr. BAESLER. Did you combine 11 studies into one big study, or is that wrong? Mr. BAYARD. That is correct. We used 30 studies and when you use a meta-analysis you decide whether or not the studies are dif- ferent between countries-the results are different between coun- tries. We found that we had eight countries which broke into six country groupings. There were 11 United States studies and 5 from Japan and 4 from China, Greece, the United Kingdom, and Sweden. We found that the results differed between countries. Mr. BAESLER. The 11 U.S. studies, that is where the term 11 comes from? Mr. BAYARD. That is correct. Mr. BAESLER. How many of those studies had concluded there wasn't a problem, of the 11? Mr. BAYARD. Out of the 11, probably none, but let me tell you how. Out of those 11, 8 showed increased risk. Between one and SVTSMVOz

r 22 three was statistically significant, depending on how you count them. One showed a slightly decreased risk. That was Janerich, which found a highly significant increased risk for childhood exposure. I am trying to remember the two that actually showed decreased risk, and I don't remember offhand. Mr. BAESLER. Eight of the studies showed increased risk and you said according to how you view it, it could be a problem? Mr. BAYARD. I am sorry, I missed the question. I don't under- stand you. Mr. BAESLER. Not necessarily one to three. I thought you had a statement about according to how you view it could have been a problem. Mr. BAYARD. Eight showed increased risks for the ever versus never exposed. One of those was statistically significant for the ever versus never exposed. Mr. BAESLER. One of the eight? Mr. BAYARD. That is correct. Another two were statistically sig- nificant if you looked at dose response trends or the high exposure groups. Mr. BAESLER. Let me ask a simple question. You individually, did you think there was aproblem when you started? Mr. BAYARD. No. I didn t think that at all. I started in 1988. I didn't know what an RSP was. Mr. BAESLER. You said you started in 1988- Mr. BAYARD. I first was introduced to this problem in 1988. All I do is risk assessment for a living. That is my job. This is just an- other pretty face. I do these things. I don't belong to any program office. We are a group that does this for a living. That is my train- ing. So my answer was no, when I first started I didn't believe it at all. It was only when I saw the evidence on dose response trends and the epidemiology studies that I couldn't explain any other way. Mr. BAESLER. Thank you. Mr. ROSE. Mr. Goodlatte from Virginia. Mr. GOODLA'rrE. I am concerned about how this environmental tobacco smoke policy guide was developed. Can you-Dr. Farland, can you enlighten me on that? Mr. FARLAND. Mr. Goodlatte, the policy guide is a product of the Office of Air and Radiation, not of our Office of Research and De- velopment. We had no involvement in the policy guide other than to make sure that they didn't somehow change the science that was being provided to them through our report. Mr. GooDLATrE. So you don't know how they contracted for the development of that guide? Mr. FARLAND. That was not within our purview. Mr. GoODL.A'rrE. Is there anybody here with you today that- Mr. FARLAND. They are in the other hearing. Mr. GoODLATTE. Maybe I need to go to the other hearing. Mr. BAYARD. You could have invited them over. Mr. GOODLATTE. Let me go back to this 95 percent confidence level versus 90 percent confidence level, Dr. Bayard, do you under- stand the implications of this study? I 23 Tell me what you understand are the public implications of the study that you put out here? Mr. BAYARD. Of the EPA study? Mr. GOODLATTE. Yes. Mr. BAYARD. It hasn't been helpful to me. Mr. GoODLArrE. I am talking about the enormity of this; not just talking about the effect on a major industry , but the implications and considerations every person has to take into account based upon this report that you presented. This is an enormous consider- ation, what your association is with others in the workplace, your homes, your children, and so on. Mr. BAYARD. Much more than I ever thought it would be. Mr. GoODLnTTE. Under those circumstances, do you think it is appropriate to use a lower standard to measure the tests than you use ordinarily in epidemiological studies? Mr. BAYARD. Let me turn that around and say if I am testing at true environmental levels where everyone is exposed, do I want to be 95 percent certain that something causes cancer or am I happy to be 90 percent certain? Mr. GoODLA'rTE. I don't think that is the measure here. The measure here is the number of tests that you can throw out at the 95 percent level as compared to the 90 percent level. Mr. BAYARD. If the question is what is the difference in signifi- cance tests as between the one-tailed and the two-tailed- Mr. GoODLATTE. Here we have what could be one of the most im- portant studies that you have ever participated in, and one of the most important considerations that this panel will consider regard- ing the danger of something to society, and you step down to a lower level of confidence, and I don't understand why you do that. The Washington Post on June 23, quoted EPA as saying that the unquestionable link between smoking and lung cancer makes it de- fensible to accept a lower standard of proof in the case of ETS. Now, we don't accept that kind of lower standard in other meas- ures of determining culpability in our society. We don't say that if somebody is guilty of one murder and it is proved beyond a reasonable doubt that we can accept a lower standard, that we can accept a lower standard of significantly like- ly because they have already been convicted of one murder if the other murder is unrelated to that-we don't accept that in terms of scientific possibility, sayin~ the first poll showed that x was like- ly to be the opinion, so we will accept a lower standard now. Since we are doing it again, and the first one turned out that way we expect the second will turn out that way. Mr. BAYARD. Do you willingly expose yourself to asbestos because it happens to cause cancer in workers exposed somewhere around 100 times what you might get from a little dose? Mr. GOODLATTE. The studies conducted should be accepted. When there are other studies out there that countervail that, why would you apply a lower standard? Mr. BAYARD. There were 30 studies out there and we tried to in- clude them all. I think that is something that wc did that hasn't been done as much in the past. In the past EPA reports have fo- cused more on the positive studies. We tried to focus more on the negative studies. GV M{'.7N09