Tobacco Documents Online

I I a I I A Critical Examination of the OSHA ETS Risk Assessment T.D. Sterling, Professor W.L. Rosenbe,uin, Senior Research Associate ... ....... . J. J. lVeinkain, Professor Faculty of Applied Sciences, School of Computing Science Simon Fraser University Burnaby, British Columbia, Canada, V5A lS6 July 27, 1994

Contents .... . 1 Executive Summary 4 2 Introduction 8 3 Methods 7 3.1 Notation ..................................... 7 ' I 4 I t 5 I I I I 6 7 I I I I 3.2 Outline of the Mathematical Model . . . . . . . . . . . . . . . . . . . . . . 8 OSHA's Risk Estimates 11 4.1 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . ... . 11 4.1.1 The Proportion of Employed Non-Smokers With Workplace ETS Ex- p osure . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . 12 4.1.2 The Relative Risk of Workplace ETS Exposure Among Non-Smokers 12 4.1.3 The Mortality Rates of Employed Non-Smokers . . . . . . . . . . 13 Result of OSHA's Risks Assessment 14 An Evaluation of OSHA's Procedure 15 6.1 The Mathematical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 6.2 How Appropriate Are the Population Parameter Estimates Utilized in OSHA's -- Mathematical Model? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 6.2.1 The Proportion of Employed Non-Smokers Exposed to ETS in the Workplace . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . 18 6.2.2 The Relative Risk of Lung Cancer and Heart Disease of Employed Non-Smokers With Workplace ETS Exposure Versus Employed Non- Smokers with no Such Exposure . . . . . . . . . . . . . . . . . . . . 20 6.2.3 The Mortality Rate of Employed Non-Smokers . . . . . . . . . . . . 27 What Can Validly Be Done to Estimate Lifetime Lung Cancer Mortality Risks Associated With Workplace ETS Exposure? 28 7.1 Data Sources . . . . . . . ; :. ' . . . . . . . . . . . . . . . . . . 7.1.1 Using the NMFS and the NHIS to Estimate Mortality Rates .... 7.1.2 The Estimation of Relative Lung Cancer Risk Associated With Work- place ETS Exposure Using Meta-Analysis . . . . . . . . . . . . . . 7.2 Results of a Risk Assessment for ETS Exposure in the Workplace ..... 29 29 30 31 8 What Can Validly Be Done to Estimate Lifetime Heart Disease Mortality Risks Associated With Workplace ETS Exposure? 34 9 Discussion 34 ~ 0 ~ ~ I

I I I I I ~ I I I I ~ I I List of Tables I Lifetime Occupational Risk for Non-Smoking Workers Exposed to ETS•in the Workplace - OSHA Estimates (Cases per 1000 workers at risk) ..... 14 '2 Percentage of Employed Non-Smokers Aged 20 to 64 by Various Confounding Variables for Persons With and Without Workplace ETS Exposure .... 17 3 Reported Relative Lung Cancer Risks of Workplace ETS Exposure .... 21 4 Lifetime Risk - Number of Lung Cancer Deaths per 1000 Workplace Exposed Non-Smokers Based on NMFS/NHIS Employed Non-Smoker Mortality Rates 33 List of Figures I Reported Relative Lung Cancer. Risks and 95% Confidence Intervals Asso- ciated N-Vith Workplace ETS Exposure . . . . . . . . . . . . . . . . . . . . 23 2 Reported Relative Lung Cancer Risks and 95% Confidence Intervals Asso- ciated ~Vith ~-~Iorkplace ETS Exposure. Studies Reporting on US Data .. 23

Abbreviations I I I I I I I ii CPS Cancer Prevention Survey EPA Environmental Protection Agency .............. . . ........ ,. ~ _. ETS Environmental tobacco smoke NHIS National Health Interview Survey NMFS National Mortality Followback Survey OSHA Occupational Safety and Health Administration RR Relative risk 3

1 Executive Summary I I I I In response to a recent request for information on indoor air quality problems the US Oc- cupational Health and Safety Administration (OSHA) has proposed a standard addressing indoor air quality in general, and especially environmental tobacco smoke (ETS), in in- door work environments. As justification for their standard, OSHA relies on a quantitative risk assessment used to provide estimates of lifetime risk of lung cancer and heart disease associated with workplace exposure to ETS. However, there are a number of concerns regarding the OSHA risk assessment. . The form of the underlying mathematical model used in the risk assessment is in- appropriate. For example, no account is taken for differences in race, age, and sex between those exposed to ETS, and those not so exposed. Workp ace exposure to ETS is assumed to be the sole cause of any difference in mortality rates between the exposed and unexposed, . OSHA was highly selective in choosing what data values to use in their risk assess- ment. ment. For example, OSHA used only the Fontham et al (6] study to estimate relative lung cancer risk associated with ETS exposure. In fact several studies meet OSHA's acceptance criteria_and provide alternative estimates of the relative risk. s Many data values required as input to the OSHA risk assessment model are simply not known at this time. When such values are required, known, but possibly inappropriate values were substituted. For example, where an estimate of relative heart disease 4 I I

risk associated with workplace ETS exposure is required, an estimate based on ETS exposure within the home is substituted. Since ETS exposure within the home is highly correlated with socioeconomic status and with paraoccupational exposure of the home members, such a substitution may be inappropriate. I I t I I 1 I I We present an alternative risk assessment, avoiding the above problems with the OSHA risk assessment. Data from the National Mortality Followback Survey and the National Health Interview Survey were used to obtain nationally representative never-smoker lung cancer rates. Meta-analysis was employed to make use of all available data regarding lung cancer risk associated with workplace ETS exposure. To the extent that these data allow, partial control for confounding from age, race and sex was performed. The results of this risk analysis show no elevated lifetime lung cancer risk associated with workplace ETS exposure. Insufficient dataa exists to perform a similar analysis for heart disease. ~ ~ 5 W 4 ~ . ~

I ~ I ~ I I I I I I 2 Introduction In response to a recent request for information on indoor air quality problems the US Oc- cupational Health and Safety Administration (OSHA) has proposed a standard addressing indoor air'quality in indoor work environments. The provisions of the proposed standard will apply to all indoor "nonindustrial work environments". A large proportion of the OSHA standard deals with workers' exposure to environmental tobacco smoke (ETS). As justification for their standard, OSHA relies on a quantitative risk assessment used to pro- vide estimates of lifetime risk of lung cancer and heart disease associated with workplace exposure to ETS (cf Section IV. of the OSHA report [20, pp.15992-16000]). Our critical review will address three issues: 1. Just how did OSHA proceed in estimating risks associated with workplace ETS ex- posure? Which model did they use? How was adjustment made for key variables that were either included or omitted? The purpose of this first discussion is to simply clarify the method, the mathematical model and the data used by OSHA to arrive at their risk estimate. 2. Is the risk analysis performed by OSHA valid? If not, why is it not valid? 3. How could presently available data be used to assess the risk of ETS in the workplace? 6

I I i 3 Methods 3.1. Notation We use the following notation, consistent with that developed in [24]. Quantities of interest I include the following: • P - persons at risk • D - deaths occuring among the population at risk • R - annual mortality rate • RR - relative risk of workplace ETS exposure . Q - proportion of persons exposed to ETS in the workplace . ED - excess deaths due to workplace ETS exposure . AR - attributable risk due to workplace ETS exposure . LAR - lifetime attributable risk due to to workplace ETS exposure These quantities may be indexed by the following subscripts. 9 c - cause of death (Lung Cancer,Heart Disease) • s-sex s r - race . a - age . k - smoking status (k = non-smoker) 9 e - employment status (employed,not employed) s x - workplace ETS exposure status (a = not exposed) 7 I

The absence of a subscript indicates that the corresponding variable is ignored in com- I I I 1 puting the quantity. Thus for example, R,,,.ak,s refers to the annual mortality rate for cause c ainong employed non-smokers of sex s, race r, age a, who are exposed to ETS at their workplace. The above quantities are unknown population parameters. They are estimated from various data sources, including census data, special purpose surveys and epidemiological studies. An estimator for a population parameter is indicated by putting a" over the parameter, as in k,,.axes. 3.2 Outline of the Mathematical Model The parameter of major interest is the lifetime attributable risk due to workplace ETS exposure. In turn, this quantity depends on the number of deaths which may be ascribed to workplace ETS exposure. The number of deaths from cause c due to workplace ETS exposure for persons of sex s, race r, age a, smoking group k, and employment status e is calculated using the "subtractive method" (24). EDearaker = Dcsrake - ParakeRc:rake.e Psrake (• icarake - Rearaket) (1) Thus, the number of excess deaths due to exposure to ETS is obtained by subtracting from the number of observed deaths (DeJrake) the number of deaths that would have occurred 8

I I I I I ~ ~ I I I had the persons at risk been subject to the mortality rate of those persons not exposed to ETS in the workplace (PsrakeResraxef). No data exist for estimating the absolute mortality risk among those not exposed to ETS in the workplace (Resraket). However, the need for these data can be eliminated, since P Dcsrake Rcsrake - P.rake Desraker Psraker Desrakez Psrake= Psraker Psrake PsrakeY Psrake = RcsrakexQsraker + Rcsrake3(1 - Qsraker) so that Thus Rcsrake RRcsral:exQsraker + (1 - Qsraker) This equation expresses the mortality rate among those not exposed to ETS in terms of the mortality rate of the population at risk (Resrake), the relative risk of exposure to ETS (RReJraker) and the proportion of the population at risk exposed to ETS (Qsrakes)• Substituting this equation into equation I gives the formula Qsraker (RRcsrakex - ~ ~ %Dcsrnket = ~esrake [/lsrakea (RResraker - 1) + 1 Rcsrnke _ RRcarakerQsral:er + (1 - Qsrakes) RcsraL'eS Rcsrakei ~--- - I

The attributable risk among the exposed is then given by I ARcarakex _ EDesrakes1Psrake = Rurake ~n~ [Qsrakex (Ri ~esrakes ~ 1) + 1 (2) I I I I I I I t I ARc,rake= is interpreted as the annual probability that a persons of sex s, race r, age a, smoking status k and employment status e will die from cause c as a result of workplace ETS exposure. Over the course of a working lifetime (ages 20-65) the probability of not dying due to cause c as the result of ETS exposure is jIQS 20(1 - ARc,rakex). Thus the (working) lifetime occupational risk is given by 65 LARcsrxex = 1 - 11 (1 - AR,ake=) a_zo (3) ~ ~ ~ Gn 10 '~ ~ IA to

I I I I I I I I I I ~ I I 4 OSHA's Risk Estimates OSHA did not consider differences in sex, race, age and smoking status between the exposed and unexposed groups; accordingly the corresponding subscripts do not appear in the formulas they used. In particular, the attributable risk among the exposed then becomes: AR{AEr = ED~k-.z/PLe = R~~~ ~Q~s (RR~R=c 1) + 1~ (4) Since this quantity ignores age, the lifetime occupational risk is obtained from equation 3 as LARrx= = I - (1 - AR.ck= )4s 4.1 Data Sources (5) The population parameters used in equation 3 are in general unknown, and must be esti- mated from a variety of sources. Specifically, there are three sets of parameters that must be estimated. • Ql.,s The proportion of all employed non-smokers who are exposed to ETS in the workplace. . RR~~,.., The relative cause. c risk of workplace ETS exposure among employed non- smokers.

+ R~k., The cause c mortality rates among employed non-smokers. 4.1.1 The Proportion of Employed Non-Smokers With Workplace ETS Expo- sure I I I I ~ I The OSHA risk model assumes a constant proportion of employed non-smokers are exposed to ETS in the workplace, regardless of sex, race, age or non-srnoking status (ie never smoker or former smoker). They offer two estimates of this constant proportion: • 0.1881 - from the 1991 National Health Interview Survey (NHIS) Health Promotion and Disease Prevention Supplement [7]. • 0.4867 - from Cummings et al (3]. 4.1.2 The Relative Risk of Workplace ETS Exposure Among Non-Smokers OSHA restricts its. risk assessment to two causes of death: lung cancer and heart disease. The OSHA risk model tacitly assumes the relative risk comparing employed non-smokers with workplace exposure to ETS to employed non-smokers with no such exposure is inde- pendent of sex, race and age differences in the groups being compared. For lung cancer, OSHA uses an estimate of RR(Iu„9)I-,= = 1.34 for all sexes, races, and ages, based on Fontliam et al [6]. For heart disease, OSHA uses an estimate of RRlhenrtliiu = 1.28 for all sexes, races, and ~ ages, based on Helsing et al [10]. ~ ~ 12 CZ) W ~ ~

4.1.3 The Mortality Rates of Employed Non-Smokers I I I I I I I I I I I I I As with the relative risk parameter, the OSHA risk model assumes that the annual mortality rate of lung cancer and heart disease for employed non-smokers is independent of sex, race and age. For lung cancer, OSHA uses 0.121 deaths per 1000 non-smokers, as estimated from the Cancer Prevention Survey (CPS) conducted by the American Cancer Society. For heart disease, the Framingham study provides an estimate of 3 deaths per 1000 persons aged 35 to 64. ~ ~ 13 on tU ~ I

5 Result of OSHA's Risks Assessment I I I I I I t ~ I In the OSHA model, each estimate of lifetime risk depends upon estimates of three pa- rameters: the proportion of non-smoking workers exposed to ETS in the workplace, the niortality rate of employed non-smokers and the relative risk of non-smoking workers com- pared to non-smoking workers with no such exposure. For both lung cancer and heart disease two estimates of the proportion of non-smoking workers exposed to ETS in the workplace are used: the "low proportion exposed" estimate (0.1881) from the NHIS and the "high proportion exposed" estimate (0.4867) from Cummings et al [3]. For lung cancer, the mortality rate estimate of .121 deaths per 1000 persons, from the CPS and the relative risk estimate of 1.34 from Fontham et al [6] are used. For heart disease, the mortality rate of 3 deaths per 1000 persons from the Franiingham study and the relative risk estimate of 1.28 based on Helsing et al [10] are used. Table 1: Lifetime Occupational.. Risk for Non-Smoking Workers Exposed to ETS in the Workplace - OSHA Estimates (Cases per 1000 workers at risk) Proportion Exposed Lung Cancer Heart Disease Low (0.1881) 0.4 7 High (0.4867) 1.0 16 The results of OSHA's risk assessment, in terms of lifetime risk (number of deaths per thousand exposed non-smoking workers) are given in Table 1. For the "low proportion" of exposed non-smokers oS.HA estimates a lifetime lung cancer risk per thousand of 0.4 for lung cancer, 7.0 for heart disease. For the "high proportion" of exposed non-smokers, the 14

a I U I I I I I I ~ I I I I lifetime risk estimates are 1.0 for lung cancer and 16.0 for heart disease. 6 An Evaluation of OSHA's Procedure , 6.1 The Mathematical Model The subtractive method appears to be a simple ineans of assessing the impact of expo- sure to an agent. However, this apparent simplicity belies two very strong, implicit and untestable assumptions that must be made to ensure the numbers produced by the model are meaningful. Specifically, the subtractive method assumes that . Persons with and without workplace ETS exposure do not differ with respect to confounding risk factors that are not explicitly controlled for. . Workplace exposure to ETS is the cause of the difference in mortality rates between the exposed and unexposed. In the current risk assessment, this amounts to assuming that persons with and without workplace ETS exposure differ only with respect to that exposure and do not differ with regard to other confounding variables, including for example, sex, race, age, non-smoking status (never vs. former smoker), spousal ETS exposure, nutritional habits, occupation and income. This assumption is inherent in the model used by OSHA and cannot be eliminated ~ or avoided without adding appropriate parameters to the model. ~ r .. GID 15 ~ ~ ~ I

The OSHA risk assessment lumps together males and females, whites and blacks and all age groups. Failing to adjust for age alone should be cause for concern, since mortality rates are so strongly age-dependent. While OSHA recognizes the important differences among these exposed groups I i I I I I I I ; However, the extent of absorption, distribution, retention and meta o ism o [ETS] in the body depends upon various physiological and pharmacokinetic pa- rameters that are influenced by gender,race,age and smoking habits of the ex- posed individuals. These parameters and others may result in differences in susceptibilty among ezposed subpopulations. [20, page 15974] OSHA's risk assessment does not. Failing to control for differences in confounding variables between the exposed and unexposed groups does not impede OSHA's ability to perform calculations, but it does raise serious questions about the validity and the interpretation of the results of such calculations. Data from Cummings et al [3] referenced in the OSHA report itself [20, Table IV-9, page 159951 may be used to illustrate this point. The Cummings data show that among employed males with workplace ETS exposure, 37.5% are also exposed at home. In contrast, among employed males without workplace ETS exposure, only 30.7% are exposed at home. Thus the probability of being exposed to ETS at home is 22% greater for employed males with workplace ETS exposure than for employed males not exposed to ETS in the workplace. The result of the simplifying assumption would be to overestimate whatever effects may be associated with ETS exposure. 16 I

I I I I I I I I I I Data from on the 1991 NHIS Health Promotion and Disease Prevention Supplement further show the differences in confounding variables between persons exposed to ETS at the workplace and those not so exposed. Table 2 shows the percentage of employed non-smokers aged 20 to 64 by various confounding variables for those persons with and r without workplace ETS exposure. Thus for example, 49% of non-smokers with workplace ETS exposure are females, whereas 57% of non-smokers without workplace ETS are females. The table shows that non-sivokers exposed to ETS at the workplace contain relatively more younger persons and persons exposed to ETS at home and contain relatively fewer females, persons with professional or managerial jobs, persons with family income at least $50,000, college graduates and never smokers. Again, failure to control for these confounders would result in overestimation of any risk associated with workplace ETS exposure. Table 2: Percentage of Employed Non-Smokers Aged 20 to 64 by Various Confounding Variables for Persons With and Without Workplace ETS Exposure Confounder . Workplace ETS No Workplace ETS Female 49% 57% Age 20 to 35 45% 40% Home ETS Exposure 22% 15% Professional/Managerial 36% 48% $50,000 Family Income 33% 39% College Grad 30% 46% Never Smoker 66% 72% I N ~ ~ I ~ 17 ~ I ~ Q~

6.2 How Appropriate Are the Population Parameter Estimates I I I I I I I I I I I I I I Utilized in OSHA's Mathematical Model? 6.2.1 The Proportion of Employed Non-Smokers Exposed to ETS in the Work- 0 place OSHA's risk model assumes a constant proportion of employed non-smokers are exposed to ETS in the workplace irrespective of their sex, race, age or non-smoking status (ie never or former smokers). Two estimates for that proportion are used. One of them (.1881) is based on the 1991 NHIS, while the other (.4867) is derived from Cummings et al. The US National Center for Health Statistics has for over thirty years conducted an annual large-scale survey known as the National Health Interview Survey (NHIS). This survey is based on a probability sample of the civilian, noninstitutionalized population and is designed to provide accurate and nationally representative estimates of a large number of health related variables. More than 40,000 persons were interviewed in the 1991 NHIS Health Promotion and Disease Prevention Supplement. Cummings et al [3] drew their sample from persons who attended the Roswell Park Memorial Institute Cancer Screening Clinic in New York for a free cancer checkup during 1986. Following the checkup, persons were asked if they would participate in a study on ETS which required them to make certain observations during four days and participate in a number of interviews regarding their notes.. About 30% of the clinic attendees declined to participate in the ETS study. 18 I

Although OSHA does carry out its calculations using both the NHIS and the Cum- I I I I I I I I I I I I mings et al estimates, in the end it is only the Cummings et al estimate is given any . credence. It is difficult to see how these two estimates can be equated. One estimate is based on a 40,000 strong probability sample of the US population; the other is based on a group of individuals who were sufficiently concerned with their health to attend a nearby screening clinic, take a cancer checkup, and who furthermore agreed to become subjects in a study of ETS. In esssence, OSHA rejects the findings by NHIS and gives greater credence to the results of a small self-selected sample of volunteers. Furthermore, there is the question of consistency of the NHIS and the Cummings es- timates. Whereas the Cummings estimate is an estimate of the proportion of employed non-smokers exposed to ETS at work but not at home, the NHIS estimate is an estimate of the proportion of employed non-smokers with workplace exposure, regardless of their ETS exposure_at home. The estimate of.the proportion of employed non-smokers with workplace exposure regardless of their ETS exposure at home based on the Cummings data is 0.7789, or over 4 times as high as the tTHIS estimate. It seems clear that the two estimates cannot be measuring the same quantity. The acceptance of Cummings et al as a serious estimate of the proportion of employed non-smokers exposed to ETS is not unimportant. Estimates of the lifetime occupational risk attributable to workplace ETS exposure depend on the number of non-smokers so ex- posed. The larger the estimated proportion, the larger the calculated lifetime occupational risk. The use of the Cummings et al result is, from a statistical viewpoint, simply not an 19 I

acceptable estimate vis a vis the NHIS study. I I I I I I I ~ I I I I I 6.2.2 The Relative Risk of Lung Cancer and Heart Disease of Employed Non- Smokers With Workplace ETS Exposure Versus Employed Non-Smokers ' with no Such Exposure For lung cancer, OSHA uses an estimated RR of 1.34 and for heart disease OSHA uses an estimated RR of 1.28 for all sexes, races, and ages based on the Fontham et al and Helsing et al studies, respectively. For different reasons neither of these relative risk estimates is appropriate. Lung Cancer Table 3 presents results from twelve studies reporting relative lung cancer risk associated with workplace ETS exposure [6, 8, 11, 12, 13, 14, 16, 21, 26, 27, 29, 30]. Where studies present estimates for males and females separately, both estimates appear in the table. Only three of these risks (from Fontham et al [6] and from Wu-Williams et al [30] for females and from Kabat and Wynder (12] for males) achieve borderline statistical significance. Figure 1 presents the relative risks and 95% confidence intervals reported in these stud- ies, while Figure 2 presents results based on US studies only. The reported RR's show considerable variablity. The studies with the highest relative risk are those with the largest confidence intervals. This pattern by itself is suggestive of publication bias (ie a preferred selection for publication of studies with positive rather than negative findings) [1, 4]. How appropriate it is to single out Fontham et aTs estimate as representative of the 20 I

Table 3: Reported Relative Lung Cancer Risks of Workplace ETS Exposure I I I I I I I I I I I I 1 Study Sex Country Workplace Exposure RR 95% CI Butler (2) Male USA Worked 1-30 years with a smoker 1.72 (0.33,9.04) Butler (2] Female USA Worked 1-10 years with a smoker 1.03 (0.11,10.11) Fontham [6) Female USA Exposed at work 1.34 (1.03,1.73) Garfinkel (8) Female USA Exposure at work in last 25 years 0.93 (0.73,1.18) Kabat (11) Male USA Exposed at work 0.98 (0.46,2.10) Kabat [11] Female USA Exposed at work 1.00 (0.49,2.06) Kabat [12) Male USA Current exposure on a regular basis 3.27 (1.01,10.60) Kabat [12] Female USA Current exposure on a regular basis 0.68 (0.32,1.47) Kalandidi [13) Female Greece Exposed at work 1.08 (0.24,4.87) Koo (141 Female Hong Kong Exposed at work only 0.91 (0.15,5.37) Lee (16] Male England Exposed at work 1.61 (0.39,6.60) Lee [161 Female England Exposed at work 0.63 (0.17,2.33) Shirnizu [21] Female Japan Someone at workplace smokes 1.24 (0.73,2.11) Svensson [26] Female Sweden Exposed at work or at home 1.20 (0.40,2.90) Varela [27] Both USA 150 person/years smoking in workplace 0.91 (0.80,1.04) Wu [29] Female USA Exposed at work 1.30 (0.50,3.30) Wu-Williams [30] Female China Exposed at work 1.20 (1.00,1.40) Meta-Analysis All Countries 1.05 (0.97,1.14) Meta-Analysis USA Only 0.99 (0.89,1.09) relative risk associated with ETS exposure among non-snlokers? A number of choices were available as estimator for this parameter; why choose one over tlle other? Fonthaln et als estimate of 1.34 is the third highest among the thirteen RR's presented in Table 3, and is the second highest among the eight US RR's. Since the lifetime occupational cancer risk given by the model increases with the relative lung cancer risk (equations 4 and 5), the magnitude of the relative risk estimator used can have a dramatic impact on tlle results of the risk assessment. For example, had tlle results of either the Garfinkel (8] or the Varela [27] studies (both US studies and comparable in terms of size and quality with the 21 I

Fontham et al study) been used instead of the Fontham et al estimate, the OSHA risk assessment would have predicted a negative lifetime risk (ie a protective effect) associated with workplace ETS exposure! A more plausible estimate may be obtained by using the EPA's method of meta-analysis I I I I I I , based on all of the available workplace relative lung cancer risks. Such an analysis (de- scribed in_section 7.1.2) yields an estimate of 1.05 based on all studies and 0.99 based only on US studies. The results of use of these estimated RRs associated with workplace exposure are discussed in section 7.2. There are two misapplications of the Fontham analysis that require further comment. Fontham's study presents relative risks comparing never smokers exposed to ETS in the workplace to never smokers not so exposed. Without justification, OSHA uses these relative risk estimates as if they were comparing non smokers with workplace ETS exposure to non smokers without such exposure. This is tantamount to assuming that the relative risk for never smokers exposed to ETS is the same as for former smokers exposed to ETS. Second is the fact that the Fontham et al estimate is based on females only, and yet is applied to a population comprised largely of males. This fact is a further recommendation for using a relative lung cancer risk estimate based on a meta-analysis of all reported results - including both males and females. 22

Figure 1: Reported Relative Lung Cancer Risks and 95% Confidence Intervals Associated with Workplace ETS Exposure Y N _- 8?~i +.+ T 2.00 = T 1.00 .._ .. ..z-- -.~...~...F .. .. •- --~--- - 11 1 ~ 0.00 ~ s V ~ 61 ~ L CG +Y Y A j Y u. y ~ ~ ~ L7 ti 3 I Figure 2: Reported R elative Lu ng Canc er Risks and 9 5% Confi dence Interva ls j I 11,00 T Ass ociated Studie With Wor s Reporti kplace ETS Ex ng on US Dat posure a 10.00 I 9.00 8.00 I Y 7.00 N 6.00 ''J.00 d ~ 4 00 . I 3.00 2.00 1.00 I -• --. .....~. .... ..- . ..... . ....~.... ~ ... m ...... 0 0 . 0 9 ~ N ~ !V I E ~ ~. ~ ~ .~. ~ ~ ~ ~ ~ a cfl Y c ~ ~ -~ m co ~C Y ~ .0 6. Y I U. 23 I

Heart Disease Ordinarily OSHA's standard setting procedures derive from observed risk in the workplace. Studies estimating relative lung cancer risk for workplace ETS exposure are generally negative - however, such studies do exist in sufficient numbers and lend themselves to a risk analysis model. In contrast, however, there are only very limited I I I I I I I I I I I , results available for relative heart disease risk associated with workplace ETS exposure. In an analysis of data from the Multiple Risk Factor Intervention Trial (MRFIT), Svendsen et al [25] report a relative coronary heart disease mortality risk of 2.6 (95% CI 0.5-12.7) comparing male never smokers whose coworkers smoke with male never smokers whose coworkers do not smoke. Although the MRFIT is a large American prospective study, it is difficult to extrapolate the study's result to all of the US. Men were enrolled in the MRFIT if they were ranked in the upper 10 to 15%._.ozl a risk scale based on serum cholesterol concentration, diastolic blood pressure and smoking status. Since all the men in Svendsen et aPs analysis were never smokers, they must have all scored very high on the other two classification variables. In a case-control study of never-smoking Chinese women He et al [9] find a relative non- fatal coronary heart disease risk of 1.85 (CI 0.86-4.00) comparing never smoking women exposed at work to never smoking women not exposed. However, it is difficult to justify this result as being applicable to the whole of the United States, since the He et at study focuses on a single sex, a different country and a different health endpoint. Finally, Dobson et at (5) present odds ratios for heart attack or coronary death associated with passive smoking at work, based on a case-control study conducted in New South Wales, 24 I

Australia. However, Dobson et al find no evidence of increased risk due to workplace ETS I I I I I I I I I exposure, and with the exception of female former smokers, show all point estimates of worliplace risk as less than 1.0. The overall odds ratio estimate combining sexes and never . ........ ....... .... ........... ...... and former smokers based on the Dobson et al data is 0.99 (95% CI 0.72-1.38). , The available studies of the association between heart disease and workplace ETS ex- posure'are too few in number and too_ dissimilar to combine in a meta-analysis. This scarcity of data presents OSHA with a dilemma, which OSHA resolved by conclud- ing that "risk estimates calculated from studies of the general population, or of selected subgroups, such as non-smoking wives of smoking husbands, are relevant to the working non-smoking population". OSHA defends that extrapolation from household risk to appar- ently unobserved workplace risk by concluding that "it is the exposure to environmental tobacco smoke and not the environnlent in which that exposure occurs that is the important factor'' [20, page 15994]. However that last generalization may not be true at all. We have separately submitted a detailed discussion of several reasons why spousal studies may not measure the effect of exposure to ETS but instead reflect the effects of socioeconomic factors related to the presence of absence of a smoker within the household under the title An Alternative Explanation for the Apparent Elevated Relative Morta ity and Morbidity Risks of Spouses and Other Family Members of Smokers Associated with Exposure to Environmental Tobacco Smoke. Nevertheless we briefly review the reasons here: 25 I

1. Occupation and economic factors are so heavily confounded with smoking patterns that the presence of a smoking inale in the household acts as a surrogate for paraoc- I I I I I I I I I I I 1 . cupational exposure of other household.menibers. ,2. There are many observations of a firm link between paternal and spousal occupations in disease of other members of the household but especially of female spouses of industrial workers and of their children. 3. There is a high probability of exposure to carcinogens and other toxic materials brought home by other household members who are working in workplaces where they are so exposed. 4. There is a strong association between socioeconomic status and mortality. ~ 5. Wives married to non-smokers have been observed to have "healthier" lifestyles. 6. Elevated risks have been reported for non-smoking females living with smoking males but not for non-smoking males living with smoking females. 7. Studies of the effects of workplace exposure to ETS in general have been negative. Until and unless this confounding between socioeconomic factors and ETS can be re- solved there is no justification to use relative heart disease risk obtained from spousal studies for estimating relative heart disease risk due to ETS exposure in the workplace. 26 I

6.2.3 The Mortality Rate of Employed Non-Smokers I I I I I I I I I I The OSHA risk model assumes incorrectly that the mortality rate of lung cancer and heart disease for employed non-smokers is independent of sex, race and age. For lung cancer, OSHA uses an estimate of 0.121 deaths per 1000 non-smokers, from CPS. For heart disease, the Framinghani study provides an estimate of 3 deaths per 1000 persons aged 35 to 64 (computed as a weighted average of 4 deaths per 1000 for males and 2 deaths per 1000 for females.) However, this rate is applied to a population of persons aged 20 to 64. Since coronary heart disease is much less common among persons aged 20 to 34 than among persons aged 35 to 64, the Framinghani-based estimate overestimates the true parameter. Furthermore, the Framingham non-smoker mortality rates represent estimates for the entire non-smoking population, whereas what is actually required are rates for employed non-smokers. Due to the fact that the employed population is healthier than the general population (rvhich contains a segment who are too ill to work - this is the so called healthy-worker effect), an estimate of the general population non-smoker morbidity or mortality rates overestimates the employed non-smoker morbidity or mortality rates (22, 23). And indeed, an estimate of the mortality rate due to coronary heart disease among employed non-smokers aged 20 to 64 based on nationally representative data (see section 7.1.1 below for details) is 1.25 per 1000 persons for males and 0.41 per 1000 persons for females - well below the Framingham estimates OSHA used (based on all non-smokers aged 35 to 65). Since.the attributable risk is directly proportional to the employed non-smoker mortality 27 I

rate (equation 2 and 4), the result overestimates the attributable risk and hence the lifetime occupational risk. 7 What Can Validly Be Done to Estimate Lifetime Lung Cancer Mortality Risks Associated With Work- I I I I I 11 I I I place ETS Exposure? In this section a lifetime lung cancer occupational risk assessment computation similar to that done by OSHA is carried out. The differences between the computation performed here and the OSHA computation are: . The use of the NHIS and NMFS to estimate lung cancer mortality rates for employed non-smokers based on large representative probability samples of the US population and of US decedents. These estimates are used in place of the CPS estimate of non-smoker mortality. . The use of ineta-analysis to obtain an estimate of relative lung cancer risk of workplace ETS based on all available studies. This estimate is used in place of the estimate based on the single study, Fonthain et al [6]. . The adjustment for age, race, and sex described in equations 6 and 7 in section 7.2. 1 28

7.1 Data Sources 1 I 11 I I 7.1.1 Using the NMFS and the NHIS to Estimate Mortality Rates Use of the 1986 National Mortality Followback Survey [18] and the 1987 National Health Interview Survey Cancer Epidemiology Supplement [19] allows some measure for controlling .: _ for sex, race and age. The 1986 National Mortality Followback Survey selected a representative sample of 18,733 death certificates of persons who had died in the United States during 1986. In- formation on cause off death, race, sex and age came from the death certificate. Further information such as the decedent's longest held occupation, annual income and smoking habits was also obtained for each sampled decedent. _ The National Health Interview Survey collects information on a nationwide sample of households as part of the ongoing activity of the National Center for Health Statistics. Each week, a sample of households is selected from the civilian, noninstitutionalized popu- lation in such a way that the weekly sample is representative of the target population and the weekly samples are additive over time. The 1987 National Health Interview Survey sampled 49,569 households comprising a total of 122,859 individuals. The 1987 Cancer Epidemiology Supplement of the National Health Interview Survey obtained information about the smoking habits of a representative random subsaniple of 22,080 persons chosen from the full sample. Each sampled individual (decedents in the NMFS, general population in the NHIS) is on 29 W ~ ~ W ~

I I I I I I t I assigned a weighting factor which represents in essence the reciprocal of the probability that the individual was selected into the sample. By adding up the weighting factors•for all sanlpled persons with a particular characteristic of interest (for example, employed male non-smokers aged 45 to 65) it is possible to obtain an estimate of the number of persons , with that characteristic who died in the US in 1986 (from the NMFS) or who were alive in the US in 1987 (from the NHIS). The quotient of these two numbers is then a nationally representative estimate of the annual mortality rate among persons with the characteristic. Further details on using simultaneous surveys to obtain nationally representative mortality rate estimates may be found in [2S]. 7.1.2 The Estimation of Relative Lung Cancer Risk Associated With Work- place ETS Exposure Using Meta-Analysis There are twelve studies reporting relative lung cancer risk associated with estimated work- place ETS exposures (6, 8, 11, 12, 1:3, 14, 16, 21, 26, 27, 29, 30). Following a method de- scribed by Yusuf [31), estimates from each of these individual studies were assigned weights inversely proportional to their standard errors, and a weighted-average relative risk was computed. This is the same method of meta-analysis used by the EPA in its assessment of lung cancer risk associated with spousal smoking'. The resulting estimate of relative lung cancer risk associated with workplace.ETS exposure based on these twelve studies is 1.05. When the meta-analysis is limited to US studies only [6, 8, 11, 12, 27, 29] in order to 1Meta-analysis may or may not be appropriate to apply to studies of lung cancer associated with workplace ETS exposure. It is presented here as an illustrative device only. 30 I

control for differences in working and other conditions, the resulting estimate is 0.99. I I I I I I I I I 7.2 Results of a Risk Assessment for ETS Exposure in the r Workplace In recomputing the OSHA lifetime risk estimates using NMFS and NHIS data, persons were classified by sex, race (white, non white) and age (20-34, 35-49, 50-64). Attributable risk was obtained for each cause, sex, race and age group (ARts,.a;~e) and for all age groups combined (ARts,.xe). The crude lifetime occupational risk is given by LAResrke = 1 - (1 - ARrsrke)45 (6) and the age-adjusted estimate is given by ............. ........ .. . .... .... .. LAR,fTke AR~:,r(20-34)Je)15 (1 - ARzs,.(35-49)ke) 'S (1 - ARedr(SO-s5)ke)15 (7) Results of this risk assessment, expressed as number of lung cancer deaths per thousand workplace exposed non-smokers, are shown in Table 4. The table shows lifetime risk for males and females and whites and nonwhites separately as well as for all races and all sexes. Results are shown both adjusted and not adjusted for age. The column on the right shows OSHA's estimates for comparison. It should be kept in mind that OSHA's estimates are . ~ not adjusted for age. 0 ~ 31 ~ ~ ~ O

Lifetime risk adjusted for age are consistently larger than lifetime risk not so adjusted. Lifetime risk for males are larger than those for females. For all races and sexes combined, the 'risk unadjusted for age is 0.15 lung cancer deaths per thousand workplace exposed non-smokers compared to 0.40 as incorrectly estimated by OSHA. I I I I I , Using the meta-analysis based exclusively on studies performed in the US, the relative risk is slightly smaller than 1.0. As a result the estimated number of lung cancer deaths per thousand non-smoking workers exposed to ETS is less than the estimated number of lung cancer deaths among those not exposed. To advocate that ETS exposure in the workplace would decrease the ris.....k.. for lung cancer is of course pure nonsense. Rather the results reflect the assumption inherent in the mathematical model that the exposed and unexposed groups differ only in their exposure. The computations stated in equations 6 and 7 provides partial control for some, but not all, of the possible confounding risk factors. However, the results of Table 4 clearly indicate that there is no demonstrated elevated lifetime risk for ETS exposure in the workplace. 32 I

Table 4: Lifetime Risk - Number of Lung Cancer Deaths per 1000 Workplace Exposed Non Smokers Based on NMFS/NHIS Non Smoker Employed Mortality Rates ` Low Proportion Exposed - Meta-Analysis Relative Risk (1.05) Males White NonWhite All Races Adjusted 0.29 0.36 0.30 Not Adjusted 0.17 0.15 0.17 Females White NonWhite All Races 0.12 0.17 0.13 0;07 0,08 0.07 Males And Females Combined White NonWhite All Races 0.22 0.26 0.22 0:12 0:11 0.12 High Proportion Exposed - Meta-Analysis Relative Risk (1.05) Males White NonWhite All Races Adjusted 0,73 0,90 0.75 Not Adjusted 0,43 0;37 0.42 ' Females White NonWhite All Races 0,30 0,43 0,32 0.17 0.20 0.17 Males And Females Combined White NonWhite All Races 0,55 0,66 - 0.56 0.31 0.28 0,31 Low Proportion Exposed - US Meta-Analysis Relative Risk (0.99) Males White NonWhite All Races Adjusted -0,06 -0.07 -0:06 NotAdjusted -0.03 -0.03 -0:03 Females White NonWhite All Races -0.02 -0,03 -0.03 -0.01 -0.02 -0.01 Mates And Females Combined White NonWhite All Races -0.04 -0.05 -0.05 -0:02 -0.02 -0.02 High Proportion Exposed - US Meta-Analysis Relative Risk (0.99) ' Males White NonWh(te Atl Races Adjusted -0.15 -0.19 -0,15 Not Adjusted -0.09 -0.08 -0.09 Females White NonWhite All Races -0.06 -0.09 -0.07 -0.03 -0.04 -0:04 t Mates And Females Combined White NonWhtte All Races -0.11 -0,14 -0.12 -0.06 -0.06 -0.06 OSHA 0.40 OSHA 1.00 OSHA 0,40 OSHA 1.00 C. VuLs4s®(s

8 What Can Validly Be Done to Estimate Lifetime Heart Disease Mortality Risks Associated With Workplace ETS Exposure? I I I I I Insofar as there are no US studies of heart disease associated with workplace exposure based on the general population, a valid estimate of lifetime heart disease risk associated with workplace exposure to ETS can not be calculated. A valid estimate must await further studies to provide appropriate data. 9 Discussion Does the OSHA risk assessment withstand scrutiny as a scientific document? Their analysis contains a number of flaws: . There are unsupported assumption inherent in OSHA's mathematical model. The model assumes non-smokers exposed to ETS in the workplace do not differ from non- smokers _ _ smokers not exposed to ETS in the workplace with respect to any confounding factors including sex, race, age, non-smoking status, other ETS exposures, occupation, nu- tritional habits or income. The model as stated is invalid if this assumption is not satisfied. s A small, self-selected sample of volunteers from a single geographical location (ic the Cummings et al study) is preferred to the representative sample of over 40,000 34

persons from the entire US population (ie the 1991 NHIS estimate) in estimating the proportion of employed non-smokers exposed to ETS in the workplace. In the end, OSHA relied only on the highest risk estimates as based on the Cummings et al estimate. I I 1 I I i + An estimate of absolute non-smoker heart disease mortality rates among persons of age 35 to 64 is used where an estimate of the non-smoker heart disease mortality rate among persons of age 20 to 64 is called for. Insofar as persons of age 20 to 34 eexperience lower heart disease mortality rates than persons of age 35 to 64, the effect of use of this inappropriate estimator is to overestimate the lifetime heart disease occupational risk for non-smokers. s Estimates of absolute non-smoker mortality rates for the entire population are used where estimates of iiioi•tality rates for employed non-smokers are called for. Insofar as the employed population experiences lower mortality rates than the general popu- lation, lation, the effect of use of this inappropriate estimator is to overestimate the lifetime occupational risk for non-smokers. . The use of the female non-smoker lung cancer rate estimate is applied to all non- smokers. smokers. CPS is a large and well-documented study; surely estimates of non-smoker lung cancer mortality rates are available for males and females separately. For heart disease, a weighted average of male and female rates was used. No justification for the nonsynimetric treatment of lung cancer and heart disease is given. Insofar as 35 I

male non-smoker lung cancer rates are higher than female non-smoker lung cancer rates, the effect of using the female rates exclusively would be to underestimate the t I ~ lifetime lung cancer occupational risk. _ _ ..__ :.. . .... 0 s A single study (Fontham et ao is used to provide the estimate of relative lung cancer risk when several comparable studies are available for combining in a meta-analysis. (The fact that OSHA ignored the published workplace estimates while quoting nu- merous spousal studies is itself surprising.) Insofar as the Fontham estimate is much higher than the relative lung cancer risk estimated from all available studies reporting workplace risks, the effect of using the Fontham et al estimate is to overestimate the lifetime lung cancer occupation risk for non-smokers. . The Fontham et al relative risk, based solely on females, is applied to males as well. Yet there are well-known differences in lung cancer risk between the sexes. The magnitude and direction of the bias introduced by using this inappropriate estimator is unknown. . The use of Helsing's relative risk estimate requires extrapolation of the result from a single county in Maryland to the whole of the US; from the 1960's when the data were gathered to the 1990's when the result is being used; from the white-only study population to the mixture of races in the US; from persons aged 25 or more to persons aged 20 to 65 only; and from persons exposed to ETS in the home to employed non- smokers with workplace exposure to ETS. The Helsing study is a cohort study of :36

I I 1 I relatively long duration. As Lee [15] notes, smoking habits were obtained only once ... ..... ..: (at the start of the study), and there is no indication of the extent to which people might subsequently have changed both smoking habits and ETS exposure patterns. Similarly, as Mantel [17] points out, there is the possibility that persons classified as non-smokers exposed to ETS at the start of the study could themselves become smokers by the termination of the study, particularly since they lived with smokers. In surprising contrast, OSHA defends the Helsing et al study as "being generalizable o the general population" and having "controlled for misclassification to a large degree" [20, page 15995]. . Estimates of relative risk comparing never smokers with and without ETS exposure are used where estimates comparing non-smokers are called for. The magnitude and direction of the bias introduced by using these inappropriate estimators is unknown. It is within the OSHA's mandate to make rules and set standards.regulating the expo- sure of workers in the workplace. Over the years and in many disputes with industry OSHA has established an enviable reputation ainong labor unions and occupational investigators for using hard data and for their rigorous analyses. In their estimate of ETS effects, OSHA has violated just about every rule of good applied science. The object appears to have been to establish a risk at the workplace where no such risk is. ovserved. One is tempted to speak of junk science in relation to OSHA's procedures. Accepting these procedures, because not to do so may be politically 37

I I I I harmful, will cause untold damage to future risk determination. As frequent contributors to issues of importance to OSHA we can only hope that independent analysis will win out. t;iven the present state of all of the available data, no increase in lifetime lung cancer risk associated with workplace exposure to ETS has been demonstrated. The data required t for a valid assessment of the lifetime heart disease risk associated with workplace exposure to ETS does not yet exist. 38

References 11 _ [1] C.B. Begg and J.A. Berlin. Publication bias: A probleni in interpreting medical data. Journal of the Royal Statistical Society, Series B, 151:419-463, 1988. I I ~ [2] T.L. Butler. The Relationship of Passive Smoking to Various Health Outcomes Among Seventh Day Adventists in California. PbD dissertation, University of California, 1988. [3] K. Micheal Cummings, Samuel J. Markello, Martin Mahoney, Arvind K. Bhargava, Pe- ter _ _ ter D. McElroy, and James R. Marshall. Measurement of current exposure to environ- mental tobacco smoke. Archives of Eiavironmeratal Health, 45(2):74-79, March/April 1990. [4] K. Dickersin. The existence of publication bias and risk factors for its occurence. Journal of the American Medical Association, 263:1:3s5-1389, 1990. I ~ I I [5] Annette J. Dobson, Hilary M. Alexander, Richard F. Heller, and Deborah M. Lloyd. Passive smoking and the risk of heart attack or coronary death. The Medical Journal of Australia, 157:793-797, June 17 1991. [6] Elizabeth T.H. Fontham, Pelayo Correa, Anna Wu-Williams, Peggy Reynolds, Ray- mond S. Greenberg, Patricia A. Bufler, Vivien W. Chen, Peggy Boyd, Toni Alterman, Donald F. Austin, Jonathan Liff, and S. Donald Greenberg. Lung cancer in nonsmok- ing women: A multicenter case-control study. Cancer Epidemiology, Biomarkers and Prevention, 1:35-43, November/December 1991. 39 I

t I 1 I [7] National Center for Health Statistics. Data File Documentation, National Health In- terview Survey of Topics Related to Health Promotion and Disease Prevention,• 1991 (machine readable data file and documentation). National Center for Health Statistics, _ _ 1992. Hyatsville, Maryland, [8] L. Garfinkel, O. Auerbach, and L. Joubert. Involuntary smoking and lung cancer: A case control study. Journal of the National Cancer Institute, 75:463-469, 1985. [9] Y. He, H. Lam, L.S. Li, R.Y. Du, G.L. .7ia, J. Y. Huang, and J.S. Zheng. Passive smoking at work as a risk factor for coronary heart disease in Chinese women who have never smoked. British Medical Journal, 305:350-384, February 5 1994. [10] K. J. Helsing, D. P. Saiidler, G. W. Comstock, and E. Chee. Heart disease mortality in nonsmokers living with smokers.. American Journal of Epidemiology, 127(5):915-22, 1988. [11) G.C. Kabat. Epidemiologic studies of the relationship between passive smoking and lung cancer. In 1990 Winter Toxicology Forum, pages 187-199, Washington, 1990. [12) G.C. Kabat and E.L. Wynder. Lung cancer in nonsmokers. Cancer, 53:1214-1221, 1984. [131 A. Kalandidi, K. Katsouyanni, N. Voropoulou, G. Bastas, R. Saracci, and D. Tri- chopoulos. Passive smoking and diet in the etiology of lung cancer among non-smokers. Cancer Causes and Control, 1:15-21, 1990. 40 I

I i [14] L.C.. Koo, J.H. Ho, and D. Saw. Is passive smoking an added risk factor for lung .. . . .... ..... ... . ..... ......... cancer in Chinese women. Journal of Experimental and Clinical Cancer Research, ' 3(3):277-283, 1984. [],5] Peter N. Lee. Environmental Tobacco Smoke and Mortality. Karger, 1992. [16] P.N. Lee, J. Chamberlain, and M.R. Alderson. Relationship of passive smokeing to risk of lung cancer and other smoking-associated diseases. British Journal of Caracer, 54:97-105, 1986. [17] Nathan Mantel. Dubious evidence of heart and cancer deaths due to passive smoking. Journal of Cli7aical Epidemiology, 45(8):809-813, 1992. [18] NCHS. The 1986 National Mortality Followback Survey. United States National Center for Health Statistics, 1986. (Public use tapes and/or documentation may be obtained from the Followback_Su.rvey Branch, NCHS, Hyattsville, MD). [19] NCHS. The National Health Interview Survey, Cancer Epidemiology Supplement. United States National Center for Health Statistics, 1987. (Public use tapes and/or documentation may be obtained from the National Health Interview Survey Branch, NCHS, Hyattsville, MD). [20] OSHA~ 29 CFR Parts 1910, 1915, 1926 and 1928. Indoor Air Quality; Proposed Rule. Federal Register: Part 17 Department of Labor, 59(65), Tuesday April 5 1994. 41

[21] H. Shimizu, M. Morishita, K. Mizuno, T. Masuda, Y. Ogura, M. Santo, M. Nishimura, K. Kunishima, K. Karasawa, K. Nishiwaki, M. Yamamoto, S. Hisamichi, and S. Tom- ` inaga. A case-control of lung cancer in non-smoking women. Tohoku Journal of Experimental Medicine, 1.54:389-397, 1988. [221 T.D. Sterling and J.J. Weinkam. The healthy worker effect oil morbidity rates. Journal of Occupational Medicine, 27:477-482, 1985. I I ~ [23] T.D. Sterling and J.J. Weinkain. Extent, persistence and constancy of the healthy worker effect or healthy person effect by all and selected causes of death. Journal of Occupational Medicine, 2$;348-353, 1986. [24] Theodor D. Sterling, Wilfred L. Rosenbaum, and James .I. Weinkam. Risk attribution and tobacco-related deaths. American Journal of Epidemiology, 13.8(2):128-139, 1993. [25] K.H. Svendsen, L.H. Kuller, M.J. Martin, and J.K. Ockene. Effects of passive smok- ing in the multiple risk factor intervention trial. American Journal of Epidemiology, 126:783-795, 1987. [26] C. Svensson, G. Pershagen, and J. Klominek. Smoking and passive smoking in relation to lung cancer in women. Acta Oncologica, 5:623-629, 1989. [27J Luis R. Varela. Assessment of the Association Between Passive Smoking and Lung Cancer. PhD dissertation, Yale University, 1987. I ~ C3 42 '~ ~ 01' .~ ~ ~ IA,

[28] James J. Weinkam, Wilfred L. Rosenbaum, and Theodor D. Sterling. Computation of relative risk based on simultaneous surveys: An alternative to cohort and case-control ' studies. American Journal of Epidemiology, 136(6):722-729, 1992. [29] A.H. Wu, B.E. Henderson, M.C. Pike, and M.C. Yu. Smoking and other risk factors for lung cancer in women. Journal of the National Cancer Institute, 74:747-751, 1985. I I ~ I I ~ I I (30] A.H. Wu-Williams, X.D. Dai, W. Blot, Z.Y. Xu, X.W. Sun, H.P. Haio, B.J. Stone, S.F. Yu, Y.P. Feng, A.G. Ershow, J. Sun, J.F. Fraumeni, and B.E. Henderson. Lung cancer among women in north-east China. British Journal of Cancer, 62:982-987, 1990. [31] Salim Yusuf, Richard Peto, John Lewis, Rory Collins, and Peter Sleight. Beta blockade during and after myocardial infarction: An overview of the randomized trials. Progress in Cardiovasc.uiar Diseases, XXVII(5):335-371, March/April 1985. N ~ GO ~ 43 ~ ~ ~ ~