Tobacco Documents Online

I I I I I I I I I I I U I I I I OSHA DOCKET OFFICER DATE ^ =3 9 1996 TIME Comments on the Workshop Draft of Environmental Tobacco Smoke and Lung Cancer by C. Stephen Redhead and Richard E. Rowberg Comments by S. Katharine Hammond, Ph.D. Division of Environmental Health Sciences School of Public Health University of California Berkeley, California 94720 General Comments £29 The May 24 Draft of the Congressional Research Service report on Environmental Tobacco Smoke and Lung Cancer is superficial, inaccurate, and careless in the extreme. It will require extensive revision before it is of acceptable quality to be released anywhere. As the report is currently written, I would reject it if it were submitted to a peer-reviewed journal; furthermore, I would find this unacceptable work from a graduate student. This document is an embarrassment to the Library of Congress and the Congressional Research Service. As a taxpayer, I am I

shocked and deeply disturbed by this shoddy, sloppy report from the Congressional Research Service of the Library of Congress. The report reflects an inadequate understanding not only of epidemiology and risk assessment, but, more fundamentally, of science itself. Science demands a study of all the data to see how they combine to.form a picture. This is particularly true for those who are developing a new way to examine the data, itself an admirable task. In such a situation, all the literature must be reviewed to determine how well the new interpretation fits all known data; restricting analysis to only the 3 latest epidemiological studies is simply unacceptable. Although such a task is daunting, one must choose either to conduct a thorough review and reanalysis of all data, or not perform the analysis at all; anything less than a complete job is a disservice to science itself. The difficult challenge involved can be deduced by examining the extensive efforts of other prestigious organizations which have chosen to study the relationship between environmental tobacco smoke and lung cancer. The National Research Council of the National Academy of Sciences, the advisors to the United States Surgeon General, the International Agency for Research on Cancer, the National Institute for Occupational Safety and Health, the United States Environmental Protection Agency, and, Hammond Comments Page 2 I I I I I I I I I I I I I I I I I I I

I I I 1 I I I I I I I I I I I I I I most recently, the Occupational Safety and Health Administration, have each independently studied the issue with the service of several scientists working for years. In fact, one is compelled to ask why anyone would attempt to restudy this issue given the extensive analyses of the data by some of the best scientists in the world, and the fact that all these independent studies came to similar conclusions. This certainly represents a questionable duplication of effort in a world of limited resources. There clearly has been no review of the literature, which limits the authors' ability to understand and interpret the few studies they have chosen to evaluate. Only about a dozen articles from the peer-reviewed literature are cited in the entire report. Most studies of important issues commence with a decision on what type of data will be acceptable (e.g., peer reviewed literature only, anything written that is generally available in libraries, or anything that is submitted after a public call for data), and then a search is made to find all articles/talks/data which fit the criteria and are relevant. Clearly this has not been done, as extremely important, relevant data has been omitted, and the authors are overly reliant on data from the tobacco companies and those sponsored by them. This inability to find information readily available in the published literature would be unacceptable in any forum, but it is especially Hammond Comments Page 3

I distressing when members of the Library of Congress, who have at their disposal the best library in the world, do not utilize their prime resource. Finally, I must comment on the unusual course of this review. I was first told (on April 25, when I was asked to review the document) that this workshop would take place on May 22 and that the document was virtually complete already. However, the date was later changed, and the document I received on the afternoon of May 26 was dated May 24 and contained are large number of grammatical and spelling errors in addition to its scientific flaws. I had fewer than 5 business days to review the draft, and these came at a time when I had other commitments. On the other hand, the draft is so incomplete and inadequate that it does not deserve a point by point review at this point. Therefore, I am submitting comments on some of the more egregious problems, with reservations on the entire document. I feel very strongly that before this document is released, it will require extensive revision, followed by thorough internal review, further revision, and only then, a thorough external review. The document as it currently stands is too inadequate to be reviewed properly. Hammond Comments Page 4 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 I Major oints . There is no consensus among epidemiologists that relative risks or odds ratios of less than three are suspect. Biologic plausibility plays an important role in the interpretation of data, as do study design, quality of exposure assessment, and sample size. Certainly there is no a priori reason that some chemicals may not have a true risk factor between 1.1 and 3.0. . The authors demonstrate a lack of understanding of the concept of confounding. They confuse this term with effect modification, and present an incorrect definition on pages 27-28. This lack of understanding affects the entire document. Furthermore, relatively few true confounders have been identified, although researchers guard carefully against them. Other causes of lung cancer are not necessarily confounders, and the list of possible confounders on page 30 (e.g., family history of lung cancer) demonstrates their ignorance in the field of epidemiology. . The authors neglect discussion of random misclassification and its effect on the observed odds ratio. This is a major source of error in all of the epidemiological studies, and the importance of this has not been fully appreciated by the authors. Random misclassification of exposure, e.g., classifying women ~ ~ ~ Hammond Comments Page 5 EA ~ co I

married to nonsmokers as unexposed when some are definitely exposed, and classifying women married to smokers as exposed when some of them are exposed to only a relatively small amount of environmental tobacco smoke, will lead to a lower observed odds ratio than in fact exists. It is extremely difficult, probably impossible, to classify exposures completely accurately, but some studies make better attempts than others, and this must be understood in evaluating the results of these studies. The Fontham study is noteworthy in its attempt to use several different possible measures of exposures, all of which demonstrated the relationship between exposure to environmental tobacco smoke and lung cancer. However, random misclassification certainly exists within each of these analyses, as even in homes where the husband smokes one pack of cigarettes a day, the wives' home exposures to environmental tobacco smoke will vary dramatically, depending on how many cigarettes he smokes at home while his wife is at home, where his wife is in the home when he smokes, the size of the home, the size of the rooms, the ventilation, etc.; in addition, the wives will have varying exposures to environmental tobacco smoke at work, if they work outside the home, and at other locations; all of these factors will lead to a reduction of the true, underlying, odds ratio. Hammond Comments Page 6 I I I I I I I I I I I I I I I I f1l) 0 41* , ro ra cc ~ 0 cn t~ -0 I

I I I I I I I I I I I I I I I I I I . The importance of the number of subjects in a study in the resulting level of statistical significance is mentioned at one point, but most of the discussion ignores this critical factor. Thus, many of the studies which were not statistically significant had insufficient subjects to detect low level effects; others had less accurate estimates of exposure, and so had more random misclassification of exposure, which would reduce the point estimate of the odds ratio. . The authors demonstrate a lack of understanding of exposure assessment. Had they understood the elementary principles and knowledge of first year graduate students, they•would know that most environmental exposures are lognormally distributed, and they would not have been surprised that "low-level concentrations were much more_ commonly encountered than high-level concentrations" (p 20). Nor would they try to overinterpret the finding about environmental tobacco smoke exposures that "the median value is smaller than the mean (average) value" (p. 22). . At several points in the draft report the authors state the environmental tobacco smoke concentrations (or the concentrations of markers of environmental tobacco smoke) are "low." What criteria are the authors using to judge an exposure as "low?" They should carefully Hammond Comments Page 7

explain the use of such a vague term when there are no standards against which to compare measured values. Similarly, to state that "Researchers have found that the fraction of indoor RSPs attributable to cigarette smoking is typically 10 to 50 percent of the total RSPs" (page 18) is confusing the data; the homes of both smokers and nonsmokers have been combined, and average data used; actual values in particular settings may be much higher. . The "low correlation between nicotine and RSP" is found only when the data is improperly analyzed: if the concentration of RSP is plotted against that of nicotine , generally excellent correlation is found. Evaluation of the ratio of RSP to nicotine is inappropriate at concentrations under 10 micrograms per cubic meter of nicotine, i.e., when the background RSP is a significant contributor to the total RSP. . Although the authors correctly indicate the complexities of using "cigarette equivalents" (pages 19- 20), and the report acknowledges that passive smoking can "be equivalent to over 50 cigarettes' (2-1/2 packs) worth of exposure" (page 20), this point is later lost. For example, on page 22: "Comparisons of cotinine levels in smokers and nonsmokers indicate that ETS-exposed nonsmokers receive approximately 0.5 (range: 0.1-0.7) percent of the nicotine dose of an average smoker. These data have been used to estimate the cigarette equivalent Hammond Comments Page 8 I I I I I I I I I I I I I I I I ~ I I

I I I 1 I I I I I I I I I I I I dose for ETS-exposed nonsmokers." This totally ignores the statements on pages 19-20 that an evaluation based on nicotine in the given chamber experiment would lead to an estimate that "an ETS-exposed individual would breathe about one-eighth (i.e., 0.127) of a cigarette's worth of nicotine in an uninterrupted 8-hour exposure...Other compounds give far greater estimates of exposure...that passive smoking is equivalent to over 50 cigarettes'... worth of exposure." In fact, the literature also has estimates of relative exposure based on one of the carcinogens in environmental tobacco smoke, 4-aminobiphenyl; nonsmokers had 4-aminobiphenyl hemoglobin adduct levels that have been reported to be between 14 and 20% the levels found in smokers. This is important given the italicized statement on page 17, "All of the five known human carcinogens and nine probably human carcinogens in tobacco smoke are emitted at higher levels in SS than in MS." . It is totally inappropriate to examine only the upper exposure range. Once again, any evaluation of the literature must look at all the data. Furthermore, when subjects are grouped into strata of increasing exposure, the statistical evaluation is best performed not on each stratum, but rather a trend test. . Just as there is no evidence for a threshold for ~  -Dk co active smoking, there is no convincing evidence of a ~ ~ , CA Hammond Comments Page 9 ~ ~ I

threshold for passive smoking. Furthermore, there is no biological evidence for such a threshold. As in all other exposures, one would expect a dose-response effect, that is, an increased rate of lung cancer at higher doses of environmental tobacco smoke. . Extremely important information is contained within footnotes. This information sometimes totally negates the information in the paragraph which contains the footnote (e.g., footnote 37), and the paragraph should be rewritten incorporating the information in the footnote. . The expr.:ssion "background exposure to environmental tobacco smoke" is vague and highly misleading. First, within the document it has two distinct meanings: all exposure outside the home (p. 35) and occupational exposure only (p.39). Furthermore, "background" implies a common, naturally occurring level shared by all; in fact, exposure to environmental tobacco smoke in the workplace or in public places is highly variable. Indeed, the report presents data from only one study of environmental tobacco smoke, an unpublished study by R.J. Reynolds Tobacco, Inc., and Oak Ridge National Laboratory. However, there is an extensive literature reporting both personal and area sampling that has very different findings than that reported by RJR Tobacco and ORNL; that other literature should be carefully reviewed and evaluated before conclusions are drawn about where Hammond Comments Page 10 I I I I I I I I I I I I I I I I 0 ~ m ~ cn -0 w I