Tobacco Documents Online

Thus we have ample biologic evidence that passive smoking can affect the blood in ways that are injurious to the heart, and that these effects are often stronger than one would expect considering the difference in dose between active and passive smoking. To get some idea of the magnitude of the heart risk from passive smoking one must turn to the epidemiology. There are now thirteen studies7 based on over 3000 cases where either heart disease or heart death was studied relative to ETS exposure. This is an enormous data base compared to what is usually available in regulating workplace toxins. Five of the studies • 0 are U.S. based. The others come from Australia, China, England, Japan, New Zealand and Scotland. Eight of the studies are mortality studies, and four of them have data for both males and females resulting in twelve separate data points which are shown in Figure 2. The twelve odds ratios, or relative risks for the prospective studies, are plotted on a log scale against the statistical weights for each data point. As you can see, for the smaller studies there is considerable scatter, but when the statistical weight gets beyond abut 20, the odds ratios become very stable in the 1.15 to 1.30 range. The combined odds ratio for all of the studies, as shown at the right, is 1.22 and is dominated by the large, statistically significant, Helsing, et al.18 study which comprises 42% of the total cases. Many of the other studies are too small to reach statistical significance at the 95% level but they do indicate appreciable increased risks. 8 IJ 0 Oo ~ -4 CD 4~1 C) V w

• 0 OSHA has used as a lung cancer relative risk the occupational risk of 1.34 for women from the preliminary report by Fontham, et al.Z1 This is without doubt the best U.S. study and is an appropriate choice. However, the final report22 for that study is now available so the updated occupational relative risk of 1.39 with a 95% confidence interval of 1.11 to 1.74 should be used. For heart disease OSHA used the relative risk for household ETS exposure from the largest U.S. heart study, namely, Helsing et al.76 This again is appropriate. However, an alternative analysis of the data based on a somewhat different mathematical model appears in Sandler, et al.a The relative risk for males at 1.31 is the same in both papers but the relative risk for females is slightly different, namely, 1.19 in Sandler versus 1.24 in Helsing. Dr. Sandler advises that these two estimates are statistically equivalent so an average of, say, i 1.21 might be appropriate for the female relative risk. OSHA's discussion of why the household exposure data can be used for occupational exposure is appropriate and well done. Major • support for OSHA's position that there is an adverse heart effect from ETS exposure at the workplace comes from the new He, et al19 paper from Xian, China, that was mentioned earlier. This very recent study covers females only. Although the number of cases is not large, this is an excellent paper wit4 about 15 heart disease risk factors including all of the important ones either adjusted for or accounted for. These authors found a relative risk for spousal ETS exposure of 1.24, not significant but very 16

sub-classes of smokers to estimate the proportions of each category misclassified. These are then subtracted from the proportions of observed never smokers to yield the proportion of true never smokers among exposed and unexposed cases and controls. From these numbers a corrected relative risk can be calculated. Our method has been peer reviewed by EPA's Science Advisory Board. In late 1991 Peter Lee, a consultant to the tobacco industry, came up with an alternative methodaD for estimating the smoker misclassification bias. As shown in EPA's Appendix B, which is shown in condensed form in Table 2, when the same inputs are used, their method and ours give essentially the same results. However, their inputs are different so their results are different, namely, much higher bias. There are four important inputs for these calculations. They are, as shown in Table 3, (1) the misclassification rates, (2) the prevalence of smoking among the subjects (the more smokers, the more will misclassified), (3) the relative risks assumed for the misclassified smokers, and (4): the marriage concordance among smokers and non smokers, i.e., the relative degree to which smokers marry smokers and nogsmokers marry nonsmokers. Regarding the first important input, namely, misclassification rates, we used, for current smokers, all of the literature data on females where we could get individual cotinine measurements from the authors. This allowed us to tell exactly 12 T

et al studyZl on which OSSA's lung cancer risk assessment is based the correction for smoker misclassification is very small • because the authors were able to eliminate most smokers from the study by cotinine tests. of the current For heart disease there is an even smaller correction for smoker misclassification. The reason is that the relative risk for active smoking for heart disease is much lower than for lung cancer, about 1.7 versus about 4 world wide for lung cancer in females or about 8 in recent U.S. studies. This lower relative risk carries over to the smokers misclassified as never smokers, thus lowering the correction. In Wells' recent paperT on passive smoking and heart disease the pooled heart relative risk for females was lowered only from 1.25 to 1.24 by the smoker misclassification correction. For males the correction was larger, 1.38 to 1.30, because of the larger proportion of smokers. Again, these corrections are probably too large by a factor of two to four. Comments on the O8HA Preliminarv Risk Rssessment The passive smoking risk assessments that have been carried out so far both for lung cancer and for heart disease have covered the entire U.S. population. This risk assessment by OSHA is the first, to my knowledge, that is restricted entirely to workplace exposure. Although I might differ on some of the details, the assessment methodology is well conceived and the N 0 00 results are reasonable. s V 00 A O1 15 00 N

similiar to the combined result for home exposure from all of the other studies. In addition they found an almost statistically significant relative risk for workplace exposure of 1.85 with a highly statistically significant upward trencl as workplace exposure increased. As I noted earlier, Xian is a good place to study the effects of ETS exposure on the heart because very few women smoke there, about 8%, thereby cutting down on background and misclassification effects, but most of the men smoke, and they smoke freely both at home and at work. The high workplace • for significant at the 954 level, and when the risks are applied to the large number of people exposed, a very large public health risk is indicated. In the risk assessments for the total population3'S•7 it is customary to adjust the relative risks upward to account for exposure to background ETS. This is done in order to estimate the risk relative to a hypothetical population that has no ETS exposure at all. However, in OSHA's case where they are trying to determine the effect of workplace exposure alone, using a relative risk without adjusting for background is more relative risk mayy result from higher exposure than is typical U.S. workplaces, but this paper indicates strongly that ETS exposure at work can indeed result in coronary heart disease, just as the Fontham, et al paper22 suggests the same for lung cancer. Although the risk elevations from Fontham, et al.a and Helsing, et al.18 are small, the results are statistically 17 ~

how many of the false negatives among the never smokers were really regular smokers and how many were only occasional smokers. We are the only group that has gone to this extent to develop accurate current smoker misclassification rates. Similar care • was taken with the exsmoker surveys. We concluded that 1.09% of regular smokers would say never, plus 24.2% of occasionals and 11.7% of exsmokers. Lee used a single overall misclassification rate which was in effect, about 30% larger than ours. We used smoking prevalence and smoker relative risks that were appropriate for each study. Lee used values that in some cases were twice as high as indicated in the studies themselves. On marriage concordance he and we used similar values. The overall effect when these overages were all multiplied together was, that his estimated bias was in some cases several times as large as ours. There is evidence that even our estimates of the misclassification effect are too high. Our data come from some epidemiologic studies but mostly from community survey type studies. For the regular smokers and the exqmokers, the two most important classes, the misclassification rates from the community type surveys tend to run about seven times as large as those from the few epidemiolologic studies that we have. For example, Fontham, et al.,r is one of the best lung cancer epidemiologic studies. After they went through all of their normal questionnaire procedures, they ran cotinine levels on the N remaining subjects who said they never smoked. They found no 0 C* i V 00 13 ~ ~ 00 O

regular smokers misclassified versus our estimate of 1.09%, and about 10% of occasional smokers versus our estimate of 24%. Another indication that our bias estimate is high is found in the male data. Here smoking prevalences are high. When we used male misclassification rates derived as above for the female rates, we found in some studies that the number of smokers misclassified as never smokers exceeded the number of self-reported never smokers that we had to start with. This is evidently impossible, again indicating that our misclassification rates are too high. To get a measure of the impact of these corrections on the lung cancer relative risks Dr. Stewart and I pooled the relative risks from 25 of the lung cancer studies in the EPA report. The EPA had divided the epidemiologic studies on ETS exposure and lung cancer into four quality tier levels based on scores for ` various factors either included in or excluded from the various ' studies. They then based their lung cancer risk assessment on the top three quality tiers. We also used their quality assignments and drew our 25 studies from the top three tiers. These were all studies on females. We then corrected each of the studies for smoker misclassification and pooled the corrected relative risks. The pooled uncorrected risk was 1.40. The pooled corrected risk was 1.37. So, for lung cancer, the overall effect is small. For studies in western countries where female smoking prevalence is high, the corrections are larger. For Asian and other studies in conservative societies where female N O smoking is low, the corrections are very small. For the Fontham, ca v Oo A 14 0~ • 00 -a

appropriate. In the final draft some discussion of the point may be in order. OSHA may have underestimated the number of nonsmoking workers exposed to ETS at work based on the Cummings data in Table IV-9. Those exposed at work but not at home totalled 48.67% but an additional 29.22% said they were exposed both at home and at work. Therefore the range should be from 18.81% to 77.89%. OSHA used an incidence rate for lung cancer for nonsmokers. With the low survival rate this is almost equivalent to a mortality rate. However, they used a mortality rate for heart disease. Therefore the risk estimates in Table IV-lo and surrounding text are inconsistent. They are disease estimates for lung cancer and death estimates for heart disease. Since the incidence rate for heart disease is about three times the mortality rate, the true number of workers exposed to ETS who "will develop heart disease" could be about three times the numbers of 7 to 16 per thousand mentioned in the text and shown in Table IV-10. Perhaps Table IV-10 should be amended to show three cases, lung cancer incidence, heart disease deaths, and heart disease incidence. The reason that the heart disease numbers are so large relative to lung cancer is that lung cancer 4mong nonsmokers is a rare disease, so even with an appreciable increase in risk from ETS exposure, the number of lung cancer deaths from ETS exposure (about 3000 for the U.S.) is small relative to the number for N 0 ~ 18 y

heart disease (about 35,000 to 50,000). However, the ETS lung cancer deaths are still enormous when compared to those from what EPA usually regulates, such as asbestos at 15 and vinyl chloride at 27". With heart disease the underlying death or incidence rate for nonsmokers is large. The death rate is about 3/0.121 or 25 time as large as for lung cancer. So when this large rate is multiplied by even a small increase in risk from ETS exposure, very large numbers result. In conclusion OSHA is to be commended fqr quantifying these workplace risks. They are real and they need to be attended to. • 19 •

Table 1. Peer reviewed papers that have assessed the available evidence on adult mortality from passive smoking and heart disease. Wells, A.J. An Estimate of Adult Mortality in the United states from Passive smoking. Environment International 14:249-265, 1988. Glantz, S.A., Parmlay, W.W. Passive Smoking and Heart Disease; Epidemiology, Physiology, circulation 83:1-12, 1991. and Biochemistry. Steenland, K. Passive Smoking and Risk of Heart Disease. Journal of the American Medical Association 267:94-99, 1992. Taylor, A.E., Johnson, D.C., Kazemi, H. Environmental Tobacco Smoke and Cardiovascular Disease; a Position Paper from the Council on Cardiopulmonary and Critical Care, American Heart Association. Circulation 86:699-702, 1992. Wells, A.J. Passive Smoking as a Cause of Heart Disease. Journal of the American College of Cardiology 24:546-554, 1994. 689v81L80Z 3