The asbestos minerals are classified according to structural features into two groups, serpentine and amphibole. Chrysotile, a serpentine (white asbestos), comprises pliable, curl:,, fibers that are formed individually from fibrillar subunits. Layers of linked silica tetrahedra alternate with layers of magnesium hydroxide octahedra to form long, hollow, scroll-like structures. Chrysotile accounts for approximately 95 percent of the world usage of asbestos today. The major producers are the Soviet Union and Canada. The amphibole types of asbestos (crocidolite. amosite, tremolite, actinolite, and anthophyllite) are generally made up of straight, needle-like fibers consisting of strips of silica tetrahedra linked by one or more cations (calcium, sodium, magnesium, and iron). The mineral names are often distinguished by adding the modifier asbestos after the name for those minerals that may occur both as a fiber and not as a fiber. In this text, crocidolite refers to asbestiform richterite and amosite refers to asbestiform grunerite. In the United States, amosite and, to a lesser extent, crocidol;te were widely used in the past, but their commercial importance has aecreased dramati- cally in the last two decades (Craighead and Mossman 19821. The amphiboles tremolite, actinolite, and anthophy'lite are minor con- taminants of some chrysotile and industrial talc products. are present in both asbestiform and nonabestiform types, and are not produced for commercial use. The occupations and industries in which the major mortality studies of asbestos-exposed workers have been conducted are pre- sented in Table 1. Groups not described in this table, but for whom there is considerable concern about substantial asbestos exi; xxc, include workers in the building and demolition trades and mainte- nance workers. The number of workers exposed to asbestos in the United States has been variously calculated, but a detailed review by Nicholson and colleagues (1982) estimated that 18.8 million workers have had more than 2 months of exposure in occupations where significant asbestos exposure may have occurred. An earlier chapter of this Report documents that a:rf and occupation are associated with substantial differences in s Ing behavior. These differences would be expected to substantial ' ;. ter lung cancer and chronic lung disease mortality; therefore, a careful examination of the smoking habits of asbestos-exposed popula.. ,is is needed in order to interpret the data on mortality and dzw:.:`.. incidence and prevalence reported in the literature. Table 2 m-r ~2: the smoking habits of asbestos workers from a number of L' .: es of asbestos-exposed populations. In most of the studies cii ash&o<- exposed populations, approximately ; r1 to 80 percent of male asbest.cs workers smoked. In some subsets of workers, well over 9C percent of the individuals were current smokers or had smoked in the past. Li 201 157-964 0 - 8b - 8 TABLE l.-Mortality from asbestos-related diseases in various cohort studies -- `bbeat0ai.e Lung cancer w of Percent Number in TOtal MC%O- (pneum* activity Study P1tX-e Fiber type smoking cohort deaths thelioma coniceia) oteervd Expected SMR Mining McDonald et al. Quebec chrysotile 10,939 3,291 10 42 230 164 125 (1980) Nicholson et al. Quebec Chrywtile 544 118 1 26 28 11.1 252 (1979) Rubino et al. IdY Chrysotile 952 332 0 9 11' 10.4 106 (1979) Hobbs et al. WeStem Crocidolite 6,200 526 17' 14 60 36.2 157 w8o) Australia Mew-man et al. Finland Anthophyllite 66.7 1,092 248 0 13 21 12.6 167 (1974) Friction Berry and England Chryeotile M 9.113 1.640 8 NS 1435 139.5 103 materials Newhow Crocidolite ' w 4,347 346 2 NS 6' 11.3 53 w63) McDonald et al. Connecticut Chryeotile 3,641 1,267 0 12' 13 49.1 146.7 mw Crocidolite' Anthophyllite' Geneml Henderson and United Statea Cbryaotile 81 1,075 781 6 31 63' 23.3 270.4 manufacturing EIlt.ZliiP Crwidolite (1979) Ammite Newhowe and England Chryeotile M 2.887 545 46 NS 103' 43.2 238 Berry (1979) Crocidolite W 693 200 21 NS 27= 3.2 844 Amosite TABLE l.-Continued Type of activity Study Place Fiber type Aabeatosie Lung cancer Percent Number in Total MHdO- (pneumo- smoking cohort deaths thelioma conic&) Observed Expected SMR Textiles Pet0 et al. (1977) Pet0 (1980) Dement et al. (1982) McDonald et al. (1983a) McDonald et al. (1983b3 Robin et al. (1979) England England South Carolina South Carolina Pennsylvania Pennsylvania Chrysotile Crocidolite (?) Cbrysotile CmcidoIite (?) Chryeotile Crocidolite ' Chrywtile Cncidolite * Chryeotile Amoeite CrocidoMe ' Cbryeotile Amoaite CrocidoMe ' 1,106 317 10 NS 511 23.8 214 679 239 7 10 40 23.3 172 62.4 768 191 1 15 26 7.5 348 89" 2,543 857 1 21 59 29.6 199.5 75* 4,137 1,392 14 74 53 50.5 105 M 2,722 912 13 NS 49 36.1 136 w 554 128 4 NS 14 1.7 824 Cement products Weill et al. (1979) Finkelstein (1963) Thomas et al. (1982) New Orleans Scarborough CanJilT Chryaotile Crwidolite Ammite' Chrymtiie Cmcidolite Chryaotile Cnxidolite' 5,645 601 0' NS 51 49.2 104 535 138 19 NS 26 5.4 480 1,592 351 2 NS 28 33.0 85 Gas mask Jones et al. England Crocidolite 578 166 17 NS 12 6.3 190 manufacturing (1960) Insulation producta Seidman et al. New Jersey Ammite 820 528 14 30 93 22.8 408 (1979) l-3 2 TABLE l.-Continued - -- 5pe of activity Study Place Fiber type Aabeetaeie Lung cancer Percent Number in Total Me%%- (pneumw smoking cohort deaths thelioma conimie) OkrVed Expected SMR Insulators Newhouse and Berry (1979) Selikoff et al. (1979) Selikoff, Seidman et al. (1980) Shipyard Rmaiter and workers GoleE (1980) England Chrysotile I.368 83 10 NS 215 5.6 375 Ammite United States Chrysotile 82.3 17,800 2,271 175 168 486 105.6 480 and Canada Amcaite New York and Chryeotile 632 478 38 41 93 13.3 699 New Jersey Amcsite Eneland Chrvsotile 66.8 6,076 1,043 31 9 84 119.7 70 Ckidolite Amcsite NOTE: NS, not stated; M, men; W, women ' Includes one suspected cade of mesothelioma. z According to the mortality study, which WBB restricted to deaths before January 1.1978. The text of this study also noted 26 caeea of meaothelioma diagnosed to January 1.1979 ' Pleural mesotheliomas included in lung cancer total given by the authors but taken out of the lung cancer total for the purpoee of this Table `Minimal usage. n Authors stat.4 that none of the casea were clearly attributable to asbestos exposure. `Male eversmokers. 1910-1919 birth cohort. `Two c-sea did not meet witaria for entry into the cohort. `Includes mewtheliomaa and colleagues (1983) showed lower rates of smoking among shipyard workers in South Carolina. Only 42.9 percent reported that they were current smokers, and 24.8 percent had ceased smoking. This decline in smoking found in the United States is not evident in studies of asbestos workers in Great Britain. Lung Cancer Cigarette smoking is the major cause of lung cancer in the U.S. population considered as a whole (US DHHS 1982). Among U.S. men aged 50 to 70 (the group most commonly examined in occupational mortality studies), over 10 percent of the deaths were due to lung cancer in 1977 (McKay et al. 1982). The prevalence of smoking and the percentage of deaths due to lung cancer vary substantially in the studies of asbestos-exposed populations reported in the literature, but in the largest study (Hammond et al. 19791 of heavily exposed workers with a high smoking prevalence (82.3 percent), 21.4 percent of the deaths were due to lung cancer. The high incidence of lung cancer in both asbestos-exposed workers and the U.S. population, together with the potency of cigarette smoking in determining lung cancer risk, makes the determination of the smoking habits of asbestos-exposed populations essential to any evaluation of lung cancer. The prevalence of smoking varies markedly among men born in different years of this century, between blue-collar and white-collar workers (see the chapter on smoking patterns), and among the populations of asbestos workers studied in the literature. In particular, men born between 1910 and 1930 have a higher prevalence of smoking than men born earlier; men born after 1930 have had lower prevalences of smoking at any given age than the men born between 1910 and 1930. Levels of asbestos exposure have also not been constant with time. Since the recognition of the hazards of asbestos exposure, improved control of asbestos dust has reduce the levels of asbestos in mines and manufacturing plants and. more recently, in other areas where asbestos exposure may also occur. These temporal trends of smoking prevalence and asbestos dust levels result in complex relationships between cumulative asbestos dust exposure and cumulative smoking exposure. The oldest workers (those born before 1910) may have higher cumulative asbestos dust exposure at any given age than younger workers, but will have a lower smoking prevalence. Workers born between 1910 and 1930 are likely to have both a higher smoking prevalence and a higher cumulative asbestos exposure at any given age than workers born after 1930. Therefore, in many studies of currently employed asbestos workers, cumulative asbestos exposure will be somewhat correlated with smoking preva- lence, and biased estimates of dose-response relationships with 205 !z TABLE 2.-Smoking characteristics of asbestosexposed workers Number and type Study of population Smoking characteristics (percent) Comments Selikoff et al. (1968) 370 union laal members, aged 674, New Jersey Hammond et al. 17,800 union local (1979) members, New Jersey Langlands et al (1971) 252 insulation workers, Belfast Ferris et al. (1971) 183 shipyard workers Murphy et al. (1971) 101 shipyard pipecoverers. New England Harries et al. (1972) 2,443 male dockyard workers, Great Britain Harries and Lumley (1977) McMillan et al. (1979) 945 royal naval dockyard workers, Great Britain 719 royal naval shipyard workers, Great Britain SMfEX 76.5 SM' 54.2 Age <4Oyears 69' >40 years 74 SM 54.6 SM 66.4 64.7 67.2 46.7 NW 13 Ex 22.2 NS1F.X 45.4 EX 26.8 2.2 16.2 28.7 Pipe/cigar 10.5 NS 10.8 pipe/cigar 5.9 ' Never smoked regul=ly o 82.9% smoked >20 cigdday ' 19% smoked >25 cigdday NS 6.9 33.1 16.6 22.7 TABLE 2.-Continued Study Kolonel et al. (1980) Pearle (1982) Number and type of population Male shipyard workers, Hawaii 131 male shipyard workers Smoking characteristics (percent) Comments Asbestosexposed workers 63.8 Nonexpzesd workers 62.5 General population 58.8 75.6 Li et al. 3,991 shipyard workers, (1983) South Carolina SM EX NS 42.9 24.8 32.3 Be&lake et al. (1972) Asheetcm workers, Canada SM' NS 85.3 14.7 Meurman et al. (1973. 1974) Liddell et al. w32) Berry et al. (1972) Meurman et al. (1979) Anthophyllite mine workers, 1936-1967 SM' 66.7 615 asbestos workers, Bueb= 1,203 male asbestos workers Asbestos workers, Finland NS 33.8 SM Ex NS 74.5 19.5 6 Cohort survivors 66.7 Dsceasd workers 79.8 o Smokera=ever smoked 1 cig/day for 2 1 yr.; includes pipe and cigar o 26.1% smoked >15 cigslday TABLE 2.-Continued Study Number and type of population Smoking characteristics (percent) Commmts Weill et al. (1975) and Selikoff et al. (1979) 859 aebest~~~ cement mfg. SM EX NS workers, New Orleans 51 26 23 Greenberg et al. (1976) 890 ssheat-os workers, Texas 84 Weiss and Theodos (1978) 40 ashestce workers 55.7' 22.7 21.6 ' 22.7% smoked >1 pack/day Berry et al. (1979) Asbestos textile factory workers, Great Britain SM EX NS 69.2 13.8 17 Selikoff, Seidman, et al. (1980) 933 am&e asbestoe workers, examined 20 yra. from employment start date SM 61.7 EY 12.1 NS 13.4 Ottlet 12.6 Skerfving et al. 241 a&e&m workers, 64.3 ww Sweden Weiss et al. 45 asheatce workers, aged SM EX NS (1981) 240, reexamined 42.2 31.1 26.7 ~-__- TABLE 2.4ntinued Study McDermott et al. (1982) Number and type of population Two groups of a&?st.cm workers, Swaziland .- Smoking chsrxtetitica (percent) bmments SM EX NS Group 38 10 Group 3.4 4 Acheeon et al. Amosite asbestos workers. (1964) Great Britain Berry et al. 1,253 male and 423 female (1965) asbestos factory workers NOTE: SM = Smoker; EX = Exsmoker; NS = Nonsmoker. 77 5 19 Men 74.5 19.6 5.9 Women 49.4 22.7 27.9 asbestos may result. These associations between asbestos exposure and smoking must be considered when examining the literature and are particularly important when drawing conclusions from studies that either do not control for smoking or control for smoking inadequately. For these reasons, this discussion is limited largely to those studies that have provided data on the smoking habits of their populations. Examination of the relationships among smoking, asbestos expo- sure, and lung cancer includes consideration of a series of separate questions. Does asbestos exposure exert an effect in the absence of active smoking exposure? What are the effects of combined expo- sure? Is there a threshold of exposure below which no effect occurs? What happens to the risk following smoking cessation and after cessation of new asbestos exposure? Lung Cancer in Nonsmoking Asbestos Workers The most direct way to demonstrate that asbestos exposure results in an increased lung cancer risk independent of cigarette smoking is to monitor disease occurrence in asbestos-exposed individuals who have never smoked cigarettes regularly. However, because lung cancer is a relatively rare phenomenon in people who have never smoked cigarettes, even among asbestos-exposed populations, a large population of nonsmokers is required before a statistically signifr- cant number of cases would be expected. The relatively high prevalence of smoking in asbestos-exposed populations decreases even further the number of nonsmoking asbestos-exposed workers available for study, making the evaluation of risks for the nonsmok- ers difficult. For example, no lung cancer deaths were identified among the nonsmokers in the original cohort of asbestos insulation workers reported by Selikoff and colleagues (1968). Some authors have attempted to increase subject numbers in the nonsmoker category by combining ex-smokers or light smokers with never smokers (Blot et al. 1980). However, the risk of developing lung cancer remains elevated in ex-smokers compared with nonsmokers for at least 10 to 15 years after cessation, and the excess risk is proportionate to the amount smoked (US DHHS 1982). Smokers of less than 10 cigarettes per day have less risk than heavy smokers, but the relative risk for lung cancer in these light smokers compared with individuals who have never smoked regularly still varied from 2.3 to 9.5 in the major prospective studies on smoking mortality (US DHHS 1982). Thus, combining people who have never smoked with ex-smokers and light smokers is inappropriate and may introduce bias when the effects of asbestos exposure alone are being assessed. Several studies have examined populations large enough to address the question of the risk of asbestos exposure in individuals who have never smoked regularly. Hammond and colleagues (1979) 210 examined the mortality experience of the 17,800 members of the International Association of Heat and Frost Insulators and Asbestos Workers who were alive on January 1,1967. This group was followed to December 1976, and the mortality of the 12,051 workers more than 20 years after onset of exposure was analyzed. Of this group, smoking histories were available for 8,220, of whom 6,841 (83.2 percent) had been regular smokers at some point and 891 (10.8 percent) had never smoked regularly. Of the 891 workers who had never smoked regularly, death certificates indicated that 4 died of lung cancer. The expected number of deaths was calculated from the mortality experience of a population of blue-collar workers who had never smoked regularly, drawn from the American Cancer Society (ACS) prospective mortality study of 1 million men and women. The resulting expected number of lung cancer deaths of 0.7 and the observed number of 4 yielded a relative risk for asbestos exposure of 5.33. When the deaths were classified according to the best estimate of the cause of death from all available data, rather than from the death certificate alone, one additional case of lung cancer was identified in a worker who had never smoked regularly. Selikoff, Seidman, and Hammond (1980) reported the mortality of 933 men who began working in an amosite asbestos factory between June 1941 and December 1945. Of these men, 78 (8.4 percent) were known to have never smoked regularly; the death certificates of 5 of this group listed lung cancer as the cause of death. When the best estimate of cause of death was used, only three men were believed to have died of lung cancer. The expected number of deaths was 0.2, based on the ACS mortality study. This led to a relative risk of 25 (5/0.2) for workers who had never smoked regularly. McDonald and colleagues (1980) examined the mortality experi- ence of Quebec asbestos miners and millers and reported a dose- response relationship between cumulative asbestos exposure and lung cancer in nonsmokers. They compared the standardized mortal- ity ratio (SMR) for lung cancer in miners who had never smoked, using the mortality rates for the Province of Quebec, which are based on both smokers and nonsmokers. The SMR increased from 0.18 among nonsmoking miners with less than 30 million particles per cubic foot times years (mppcfoy) of exposure to 0.36 in miners with 30 to 299 mppcfey of exposure and 1.24 in nonsmoking miners with more than 300 mppcfoy of exposure. There were 19 lung cancer deaths among nonsmoking asbestos miners. These authors (McDon- ald et al. 1980) also performed a cas+control study of the 245 miners who had died of lung cancer. The distribution of cumulative asbestos exposure among the 20 nonsmoking miners with lung cancer and 20 nonsmoking control miners matched for year of birth and smoking status was examined, and the relative risk for lung cancer was found 211 to have increased from 1 in nonsmoking miners with less than 30 mppcfoy to 10 in nonsmoking miners with more than 1,000 mppcfey. Liddell and colleagues (1984) reexamined the same po~,rlation of Quebec asbestos miners after recording their smoking &tory by pack-years of exposure. They identified 223 cases of lung cancer in men who worked in the asbestos mines and mills of Quebec for a month or more before January 1967 ind who were followed to the end of 1975. The controls were selected from men in the same cohort, born in the same years as the lung cancer cases, but still living. Never smokers represented 23 of the 223 lung cancer cases and 201 of the 715 controls. The relative risks (RR) were calculated on the basis of the mortality experience of the entire asbestos-exposed population (whole population RR, l.O), and the risk in even the most heavily exposed nonsmokers was still lower than the risk in the entire population, which included both smokers and nonsmokers. The RR for lung cancer increased from 0.19 in the nonsmoking miners who had experienced a cumulative exposure of less than 100 fibers per milliliter times years ((f/mL)y) to 0.37 for those with 101 to 1,000 (f/mL)y and 0.87 for those nonsmoking miners with over 1,000 (f/mL)y, thus demonstrating a dose-response relationship with cumulative asbestos exposure for lung cancer in the workers who had never smoked regularly. Berry and colleagues (1972) conducted a retrospective study of the lung cancer mortality in more than 1,300 male and 480 female asbestos factory workers over a lo-year period and compared their mortality with the national lung cancer rates (Table 3). The national lung cancer rates were converted to smoking-specific rates by multiplying them by factors from the study of mortality of British physicians by smoking status (Doll and Hill 1964) in order to develop smoking-specific expected numbers of deaths. No lung cancer deaths were recorded among the men who had never smoked, and only one lung cancer death was recorded among the women who had never smoked. The expected number of deaths was also very low, and so even a single death was greater than expected, and it occurred in the group of women with heavy asbestos exposure. The women in the highest asbestos exposure category who had never smoked had 3.5 times the number of subject years at risk when compared with men in the same exposure category (1,404 to 399) owing to the higher prevalence of never-smoker status among women in the study. This difference in number of individuals at risk may have contributed to the demonstration of a lung cancer death among nonsmoking women but not among men. Subsequently, Berry and colleagues (1985) followed prospectively 1,253 male and 423 female asbestos factory workers from the same plants. Smoking habits were determined in 1971 at the start of the study, and the population was followed through 1980. The expected number of lung cancer deaths was 212 calculated from the death rates for England and Wales multiplied by the lung cancer SMR for greater London, and an adjustment for smoking status was made using the data from the mortality study of British physicians. Observed and expected numbers of lung cancer deaths by smoking status and level of asbestos exposure are presented in Table 4. One lung cancer death occurred among the men who had never smoked (0.1 expected) and three lung cancer deaths occurred among the nonsmoking women (0.2 expected). Meurman and colleagues (1979) reported 1 lung cancer death (of 23 total lung cancer deaths), a nonsmoking male anthophyllite miner. Acheson and colleagues (19841 also reported 1 death from lung cancer among the nonsmokers employed in an amosite manufactur- ing factory, with an expected number of 1.1. However, the expected number was calculated from age-specific population rates that included both smokers and nonsmokers rather than from the rates for a population of nonsmokers. Each of these studies supports an increased risk for lung cancer in nonsmoking asbestos workers, but the conclusions are based on a single death in a population. In summary, the evidence that asbestos exposure results in an increased lung cancer risk in the absence of cigarette smoking is based on a small number of cases, but has been confirmed in several different populations of asbestos workers. The high smoking preva- lence in asbestos workers introduces the possibility that environmen- tal tobacco smoke may increase the risk of lung cancer among the nonsmokers, particularly if the synergism demonstrated between active smoking and asbestos exposure pertains to environmental tobacco smoke as well. In spite of these concerns, the available evidence supports the conclusion that nonsmokers with substantial occupational asbestos exposure are at increased risk of developing lung cancer and that the risk increases with increasing cumulative asbestos exposure. Lung Cancer in Cigarette-Smoking Asbestos Workers The risk of lung cancer in cigarette smokers has been examined in a number of asbestos-exposed populations, and the increased risk of lung cancer in smokers, coupled with the high prevalence of smoking in many of these populations, has generated substantial numbers of lung cancer deaths for analysis. These populations differ in smoking habits, type of asbestos and duration and intensity of exposure, type of activity that resulted in exposure, and duration of the followup of the population. A number of authors have compared the lung cancer rates in asbestos-exposed populations with the rates in control populations (Table 1). This approach can establish an excess mortality in a population, but may not identify the causes of that excess. To establish a causal link between an exposure and lung cancer, specific 213 E TABLE 3.-Comparison of number of observed and expected deaths from cancers of the lung NUDlber Subjecbyeam observed lung Adjusted observed Smoking habits on of at risk Obeerved deaths cancer deaths lung cancer Expected lung January 1,lW Bubjecta (adjusted) (all cauees) WD 162, 163) deaths cancer deaths Men Low/moderate aebeatas expure Never smoked Exsmokers Smokers Not known Severe asbestos exposure Never smoked Exsmokers Smokers Not known Women Low/moderate asbestos exposure Never smoked Smokera Not known Severe a5beatoe exposure Never smoked Smokers Not known 4-i 376 2 38 335 1 509 4,423 32 219 2.122 20 41 399 11 0 0 0.0 39 415 3 2 1.6 0.2 663 6,920 82 3%5) 25.5 9.9 281 2,722 29 4 10.9 2.4 25 271 8 0 0 0.0 45 577 6 1 1 0.3 19 195 0 0 0 0.1 120 1,404 23 2w 1.7 0.2 3,474 52 19(4) 15.5 1.4 1,547 9 0 2.6 0.4 0 :2,1 0 0 0.0 0 0.1 4.6 6.2 3.4 2.0 ' Fires in parentheses indicate number of pleuralmesotheliomas. SOURCE. Berry et al (1972). TABLE I.-Observed and expected deaths from cancer of the lung during 1971-1980 Smoking habits in 1971 Number of subjects Subject-years at risk Total deaths Lung cancer deaths ObkWWd Expected' Low/moderate asbestos exposure Never smoked Exsmokem Smokenr 45 396 6 1 0.10 123 1.092 18 3 1.07 441 3.557 84 17 11.29 Severe anbestce exposure Never smoked 29 273 2 0 0.06 &smokers 123 1,003 343 8 1.25 Smokers 522 4,394 1% 35 14.63 Low/moderate asbestos exposure Never smoked Exsmokers Smokers 17 128 5 0 0.04 12 93 3 0 0.09 27 220 4 0 0.32 Severe ah&on exposure Never smoked 101 Exsmokera 84 Smokers 162 ' Calculated af?.er allowing for the effect of smoking, =x. sge, period, and region. soIJlK!E: Belly et al. (1985). 799 26 3 0.20 659 24 2 0.50 1,413 52 10 2.02 criteria must be applied to the entire body of information available on the exposure. This approach has been carefully and comprehen- sively followed for both cigarette smoking (US DHHS 1982) and asbestos exposure (Selikoff and Lee 1978), and the evidence is sufficient to establish a causal role for both of these agents in producing lung cancer. This section confines itself to an examination of their interaction. Selikoff and colleagues (1968) were the first to demonstrate increased lung cancer risk among asbestos workers in an investiga- tion that assessed smoking habits. In a group of 370 asbestos; insulation workers, none of the 48 workers who had never s: 1~ 1-9~ regularly or of the 39 workers who smoked only pipes or cigt :s developed lung cancer. Of the 283 cigarette-smoking workers, 24 died of lung cancer during the 4 years and 4 months of the followup period, although only 2.98 lung cancer deaths were expected on the basis of smoking-specific death rates. A more extensive evaluation of the risk of cigarette smor;ing for asbestos insulation workers was provided (Hammond et al. 1979) by a prospective evaluaLic,n of the 17,800 members of the International Association of Heat and Frost Insulators and Asbestos Workers discussed earlier. Of this population, 8,220 workers were more than 20 years beyond their onset of asbestos exposure and had a known smoking status. Fifty-four percent of this group were cigarette smokers at the start of the study. The comparison group was drawn from the ACS study of 1 million men and women, and consisted of 73,763 white men with no more than a high school education and not employed as farmers, but with a history of occupational exposure to dust, fumes, vapors, gases, chemicals, or radiation, who were living on January 1, 1967, and were traced thereafter. The control group was followed only until September 30, 1972, and the asbestos workers were followed through 1976; therefore, the lung cancer death rates in the control group were adjusted upward to reflect changes in the U.S. national mortality experience for lung cancer during the time period of differential fol!owup. There were 1,332 deaths among wcrkers more ihan 20 years after onset of exposure whose smoking habits were known; 314 (23.6 percent) deaths were due to lung cancer, using the best estimate of cause of death. Death certificar,e data indicated 272 lung cancer ?eai :A>, Figure 2 portt.:ya the il ,ztality ratios for smokers and :)i,nsmokerx in the col>Lrol and t.iC asbestos-exposed populations, with tne mortality raGo of nonsmokers in the control group set at 1. The lung cancer death rates increased from 11.3 per 100,000 among nonsmokers in the control group to Z3.4 in the nonsmoking asbestos workers, 122.6 for smokers in the contrG1 group, and 601.6 for smoking asbestos workers. The lung cancer relative risk with combined exposure (53.24) is far larger than the sum of the individual risks for cigarette smoking and asbestos exposure sepa- rately, and is quite close to the product of the separate mortality ratios (5.17 and 10.85) together. .\,curate data on the intensity of asbestos exposure for individual workers (dose) were not available for this group of insulation workers, .ld so an asbestos dose-response relationship was not examinea. Dosage data were available for cigarette smokers in this population, however, and the ratio of observed to expected lung cancer deaths (with the expected deaths calculated from the rates in nonsmoking non-asbestos-exposed controls) increased from 5.33 in asbestos workers who never smoked regularly to 7.02 in pipe and cigar smokers, 36.56 in ex-smokers, 50.82 in smokers of fewer than 20 cigarettes per day, and 87.36 in asbestos workers who smoked one pack or more per day. Interaction between smoking and asbestos exposure in the devel- opment of lung cancer has also been explored in other populations. In sonle studies the numbers have been too small to clearly differentiate between an additive and a multiplicative effect with combined exposure; however, the data have been consistent with an effect that is at least more than additive. This interaction of cigarette smoking and asbestos exposure has been demonstrated in asbestos factory workers (Berry et al. 1972, 19851, Quebec miners and millers (McDonald et al. 1980; Liddell et al. 19841, amosite asbestos factory workers (SeXoff, Seidman, and Hammond 1980) and Finn- ish anthophyllite miners and millers (Meurman et al. 1979). A dose-response relationship between cigarette smoking and lung cancer in the general population has been readily demonstrated in a number of prospective mortality studies (US DHHS 1982); however, dose-response relationships for asbestos exposure and lung cancer have been more difficult to establish. The carcinogenicity of asbestos may vary with the type of asbestos, and possibly with the length or diameter of the fiber. There are also potential differences in the carcinogenic risk associated with the different stages and processes of converting asbestos from the raw mineral in the mine into a finished manufactured product. As a result, it is difficult to classify the asbestos exposure of different study populations with a single measurement that quantifies the carcinogenic dose. Even if such a scale were agreed upon, actual measurements of asbestos dust levels in the work environment are often not available. Measures of dust exposures for individual workers are even less frequently available. The quantification of asbestos dust exposure has frequently used estimates of likely exposures based on work conditions and job classification, rather than actual measurements of asbestos dust in the air, because of the absence of these measurements for most workers. This lack of information has been particularly problematic for workers employed more than 20 years ago, a group now at high 217 67.36 53.24 Nonsmokers Nonsmokmg Smokers Smokrq Smoking not exposed asbestos not exposed asbestos ( > 1 pack/day) to asbestos workers to asbestos workers asbestos workers FIGURE 2.-Relative risk of dying of lung cancer for smoking and ntilsmoking asbestos workers and smoking and nonsmoking control group members SOURCE Hammond et al (19791. risk of developing lung cancer. Finally, cumulative asbestos expo- sure, age, and cumulative cigarette smoking exposure are generally 218 correlated. Older employees worked under conditions of much higher asbestos exposure than their younger counterparts, and these same older cohorts probably also had higher prevalences of cigarette smoking, as described in the chapter on smoking patterns by occupation. Confounding between cumulative asbestos exposure and cumulative cigarette smoke exposure may result when dose-re- sponse relationships between cumulative asbestos exposure and lung cancer are examined without a control for differences in smoking habits among the different asbestos exposure groups. Berry and colleagues (1972) examined dose-response relationships in a population of 1,300 male and 480 female asbestos factory workers in Great Britain. Workers were categorized as having low to moderate asbestos exposure or severe asbestos exposure, and the expected number of lung cancer deaths was calculated from stand- ardized mortality rates for lung cancer for the greater London area. An adjustment for cigarette smoking status, derived from the mortality study of British physicians by Doll and Hill (19641, was used to estimate rates for smokers and nonsmokers. The results are presented in Table 3. The small number of lung cancer deaths makes interpretation somewhat difficult, but it appears that the increased lung cancer death rate is limited to smokers with severe asbestos exposure. McDonald and colleagues (1980) examined Quebec miners and presented evidence for a dose-response relationship between cumu- lative asbestos exposure and lung cancer risk in the smoking miners. They compared the lung cancer mortality rates in the Quebec miners with the mortality rates for the Province of Quebec. Table 5 shows the SMRs for lung cancer in miners by level of cumulative asbestos exposure and smoking habits. Heavy smokers consistently had higher SMRs than moderate smokers at the same level of cumulative asbestos exposure, and the SMRs increased with increasing cumula- tive exposure to asbestos in each of the smoking categories. Using the same population of miners, these authors conducted a case- control study of 245 lung cancer victims and a similar number of control miners matched for smoking habits and year of birth. The distribution of cumulative asbestos dust exposure was examined, and the results in cigarette smoking miners showed an increase in relative risk with increasing cumulative exposure. The relative risk of cigarette smokers in the lowest exposure category (< 30 mppcfay) was set at 1.0, and the relative risk increased to 1.12 at 30 to 300 mppcfoy of exposure, 1.58 at 300 to 1,000 mppcfoy, and 1.99 at 2 1,000 mppcfoy of exposure. A more quantitative description of the smoking habits of the same Quebec miners was provided by Liddell and colleagues (1984). Their data are presented in Table 6. The dust exposure measurements were made as particles per cubic foot with midget impingers, and 219 TABLE 5.-Deaths from lung cancer in relation to dust exposure and smoking habit Dust exposure umppcf.yl accumulated to age 45 c 30 3s-299 2300 All Smoking habit 0 SMR 0 SMR 0 SMR 0 SMR Nonsmokers 5 0.18 6 0.36 a 1.24 19 0.38 Mxkratc smokers 73 1.14 64 1.35 52 2.31 189 1.41 Heavy smokers 13 2.12 11 2.39 10 4.50 34 2.63 All smoking habits 91 0.93 81 1.18 70 2.25 242 1.23 SOURCE. LIddell et al 119841 individual exposures were calculated on the basis of the work histories and the measurements of impinger dust counts in the work environment between 1949 and 1966. These counts were then converted to fibers per mL. Two hundred and twenty-three cases of lung cancer were identified and matched to 715 controls born in the same year, and a case-control analysis was conducted. As is shown in Table 6, the relative risk of developing lung cancer increases with increasing asbestos exposure category for each of the cumulative pack-year categories. The analysis also suggests that the interaction between cigarette smoking and asbestos exposure is greater than additive. Thus the studies that have examined the question of a dose- response relationship for asbestos exposure and lung cancer in the face of an adequate control for cigarette smoking have shown an increasing risk of lung cancer as asbestos exposure increases. This suggests that a dose-response relationship for asbestos exposure and lung cancer does exist, and that it is not explained by differences in smoking habits. Threshold The question whether a level of asbestos exposure exists below which an exposure does not result in an increased risk of lung cancer is one that is both technically extremely difficult to answer and extremely important to those required to make policy with regard to asbestos exposure. Current understanding of carcinogenesis and host defenses against cancer are not advanced sufficiently to allow either the acceptance or the rejection of a threshold. It is common practice to assume a linear relationship between the dose of a carcinogen and the development of carcinoma, and to assume that the dose-response relationship does not have a threshold. The linear nonthreshold model allows the extrapolation of data obtained for higher exposures 220 TABLE 6.-Risks of lung cancer, by cigarette smoking and asbestos exposure, relative to all 223 cases and 715 referents for whom smoking histories were reliable; unmatched analysis Expsure accumulated up to 9 years before death of cake mf mL'\ High and Low !&dlUrn very high Pack-years ' " 1001 ,I l.ooo1 / 2 l.ooo' All 0 Nu;nher of cases 6 7 10 23 Number of referentI 103 61 3i 201 Relative risk 0 19 0 3; n ei 0.37 1. ~ 40 Number of case5 29 2; 34 90 Number of referents 123 93 63 279 Relative risk 0.76 0 93 173 1.03 240 Number of cases 40 35 35 110 ?;umber of referents 117 79 33 235 Relative risk 1.10 1.42 2.88 1.50 All Number of cases 75 69 73 223 Xumber of referents 343 233 139 715 Relative risk 0.70 0 35 1.82 1.00 ' Number of cigarettes a day 20 x duration m pan SOURCE: Liddell et al. 119841 to the very low exposures. This extrapolation is substituted for the examination of the very large populations that would have to be examined in order to demonstrate the small expected excess risk with low dose exposure. Such models are particularly attractive for exposures for which human epidemiologic data are limited or absent. As discussed earlier, however, minimal exposure to cigarette smoke and asbestos is probably a nearly universal experience in urbanized society. Because of the large population exposed, more careful examination of the available evidence on the risks of these exposures is necessary. The number of cigarettes smoked per day by an individual is a readily available measure of the dose of smoke exposure in the active cigarette smoker; therefore, it has been possible to examine relative- ly completely the dose-response relationship for cigarette smoking and lung cancer. There is a consistent increased risk for lung cancer among smokers in the lowest category of number of cigarettes smoked per day in the major prospective mortality studies on smoking (US DHHS 1982). In the study of U.S. veterans (Kahn 19661, a relative risk for lung cancer of 3.77 was demonstrated in those who smoked only occasionally compared with those who had never smoked regularly (the relative risk for those who smoked 1 to 9 cigarettes per day was 4.07 compared with those who never smoked 221 regularly). It seems clear that for the active cigarette smoker there is no safe cigarette and no safe level of cigarette smoking (US DHHS 1982). Furthermore, recent data (IARC, in press) suggest that repetitive exposure to environmental tobacco smoke may be accom- panied by an increased risk of lung cancer, thereby suggesting that the dose-response relationship may extend even to those individuals who do not actively smoke cigarettes. The quantification of asbestos exposure is far more difficult. One method is to quantitatively estimate the number of asbestos fibers in digested lung tissue. Asbestos fibers are demonstrable in the lungs of the majority of urban dwellers (Churg and Warnock 1977); however, the number of fibers per gram of lung tissue in urban dwellers without known asbestos exposure is usually several orders of magnitude below that found in occupationally exposed workers, and the type of asbestos varies as well. Churg and Warnock (1979) assessed this urban asbestos exposure as a risk factor for lung cancer by comparing the number of asbestos bodies in 103 patients with lung cancer compared with the number in control patients matched for age, sex, smoking habits, and in some cases, occupation. No differences in the number of asbestos bodies per gram of lung tissue were found between the lung cancer patients and the control population, suggesting that, at this level of exposure, asbestos did not increase the risk of lung cancer in these patients. However, the small number of patients in this study limits the power of the study to find a small effect of asbestos lung burden on lung cancer risk. Confounding by cigarette smoking is another potential source of bias in evaluating the effects of low levels of asbestos exposure. Several of the studies presented in Table 1 do not show excess lung cancer risks at low levels of asbestos exposure, a pattern consistent with the existence of a threshold. However, lung cancer rates in the general population are determined largely by smoking habits, and if the asbestos-exposed populations have even modestly lower lifetime smoking rates, the effect of asbestos exposure may be masked. This bias is of particular importance at the relatively low levels of asbestos exposure at which the effect of cigarette smoking would be expected to predominate. Thus, in interpreting standardized mortali- ty ratios at or below 1, careful consideration must be given to confounding by the smoking habits of the workforce before conclud- ing that the levels of asbestos exposure experienced by these populations do not result in an increased lung cancer risk. In addition, modest differences in the number of cigarettes smoked per day or the age of initiation of regular smoking between the exposed population and the population from which the SMR is derived could counterbalance a modest risk due to asbestos exposure even in populations with similar smoking prevalences. 222 For lung cancer, the measurement of a threshold in epidemiologic studies is further constrained by the certainty with which the absence of an effect can be established. The precision and the accuracy of an estimation of the expected number of deaths in a workforce is heavily influenced by the detail with which the smoking behaviors are determined and the accuracy with which the lung cancer risk of a given smoking history can be estimated. In the U.S. population during 1977,lO percent of the men who died between the ages of 50 and 70 died of lung cancer (McKay et al. 1982). Therefore, a workforce with smoking patterns similar to the U.S. population would be expected to have a similar mortality experience, in the absence of any asbestos exposure. A 10 percent increase in the risk of lung cancer in a workforce (SMR 110, RR 1.1) due to asbestos exposure would mean that 1 percent of the deaths among workers aged 50 to 70 would be excess lung cancers due to asbestos, a level of risk unacceptable as the basis for an industrial hygiene standard. However, even with carefully determined smoking histories for a worksite, no data are currently available that would allow the calculation of expected death rates in smokers and nonsmokers with precision sufficient to establish that an increase of 10 percent was not simply an error in the estimates. In addition, estimates of the smoking habits of the U.S. population are not known with enough precision to adjust national or regional death rates for the smoking patterns of a given workforce so that a 10 percent difference could be considered significant. The result is a dilemma for those who would try to measure a threshold level, or an "acceptable" exposure level, for occupational exposure to asbestos: an effect too small to measure in statistical terms is still too large to be acceptable in human terms. A final caution in the determination of a threshold for lung cancer risk secondary to asbestos exposure, and in the use of such a threshold to establish environmental dust standards, is the potential differences between a threshold for lung cancer and one for mesothelioma or other asbestos-related disease. Mesothelioma, which is not associated with cigarette smoking, may occur following exposure to low levels of asbestos, and a level of dust exposure defined as a "safe" level for lung cancer risk may possibly continue to produce an increased risk of mesothelioma. A pragmatic approach to the problems of defining a threshold or establishing safe levels has been to define asbestos exposure stan- dards on the basis of the lowest level of asbestos dust exposure that can be produced with existing technology. This approach reduces the risk, but does not answer the question whether the exposure of a worker is "safe." An alternate approach has been to use the existing exposure- response data. In the face of uncertainty about the shape of the 223 exposure-response curve for asbestos exposure and lung cancer and whether a threshold exists, an assumption that asbestos has a linear exposure-response relationship with lung cancer and no threshold for effect has been suggested as both reasonable and a way to set standards (Pete 1979; NRC 1984). By definition, in this approach there can be no "safe" level of exposure (i.e., no threshold), only an "acceptable" degree of risk. However, using this method, once an "acceptable" level of lung cancer in a working population has been defined, the level of asbestos exposure that would result in that level of risk can be estimated. A corollary of this approach is that asbestos is assumed to contribute to the lung cancer that develops in populations of workers who have been exposed to asbestos regardless of their level of exposure; by extension, the asbestos found in the lungs of urban dwellers with no known occupational asbestos exposure is assumed to make a small (but finite and definable) contribution to all lung cancers. The evidence that does exist (Churg and Warnock 1977) suggests that asbestos exposure makes no "measurable" contribution to lung cancer in individuals without a definable exposure, but it is impossible to establish the absence of "any" effect. If the issues of liability can be separated from the issue of threshold, then the problem of reducing and eliminating asbestos- related disease and disability could be approached with a broader focus. The focus could be expanded beyond improving technology for reducing exposure to asbestos to include other methods of reducing the cancer risk associated with asbestos exposure. If the goal is to reduce the lung cancer deaths associated with asbestos rather than simply reducing the levels of asbestos dust in the worksite, then the deaths due to the interaction between smoking and asbestos must be dealt with, and the elimination of smoking will be a potent adjunct to environmental asbestos dust control in this task, particularly for those workers who have already received substantial asbestos exposure. A public health "feasibility" threshold could then be defined, not in terms of what dust levels were achievable, but rather in terms of what lung cancer death rates were achievable. This threshold would be the lowest cancer risk achievable, given our current technology, and would include minimizing asbestos expo- sure, maximizing smoking cessation, and applying techniques for early diagnosis and treatment. In summary, although the level of asbestos exposure that occurs in the general population does not appear to be accompanied by an increased risk of lung cancer, the demonstration of a clear threshold below which there is no effect in occupationally exposed populations is not possible. 224 TABLE `I.-Lung cancer mortality ratios with cessation of cigarette smoking in male smokers who smoked more than 20 cigarettes per day compared with those who never smoked regularly Asbestos lnsularlon workers2 10 4 115 42 34 Never smoked regularly 1 I `Data from Hammond 197!? ' Data from Hammond 19791 Cessation of Exposure A decline in the relative risk of developing lung cancer following cessation of cigarette smoking was demonstrated in cigarette-smok- ing asbestos workers by Hammond and colleagues (1979). Table 7 shows the lung cancer mortality ratios in asbestos workers who are current smokers and who have quit for varying periods of time, compared with those workers who have never smoked regularly. A companion set of numbers is provided of the relative risks for lung cancer in men not exposed to asbestos, but who are current smokers or have quit for varying periods of time, derived from the American Cancer Society study of 1 million men and women (Hammond 1972). Several authors have attempted to approach the question of the risk of lung cancer following cessation of asbestos exposure by examining the relative risks of asbestos exposure in workers following retirement (Walker 1984; Selikoff, Hammond et al. 1980). The data in Figure 3 and Table 8 reveal that the relative risk for lung cancer in asbestos workers increases and then declines with the increasing number of years from initial exposure. The workers with the longest interval from onset of exposure are also of the greatest age within the populations examined. Because of this link with age, the interpretation of this decline in relative risk as indicating that cessation of asbestos exposure results in a decline in lung cancer risk must be made with great caution. Examination of national age- specific mortality rates for lung cancer (Figure 4) also shows a decline in male lung cancer death rates with increasing age. This decline with age is an artifact of the cross-sectional nature of data 225