Document k9YNzp02pjjyprD7L3q3RX3OO

\3 } I ASSESSMENT OF RISKS POSED BY EXPOSURE TO LOW LEVELS `P ASBESTOS IN THE GENERAL ENVIRONMENT Prepared by: Marvin S. Schneiderian, Ph.D. Ian C. T. Nisbet, Ph.D. Susan M. Brett, M.P.H. Clement Associates, Inc. 1010 Wisconsin Avenue, N.W. Washington, D.C. 20007 USA Prepared for: Institut fur Wasser-, Boden- und Lufthygiene des Bundesgesundheitsantes June 1981 OOOOl UCC 000559 TABLE OF CONTENTS I.Executive Summary 1 II. Introduction 5 III. Background 7 IV. Dose-Response Data Derived from Studies of Industrial Workers 19 V.Extrapolation to Low Exposure Levels 77 VI. Gastrointestinal and Other Cancers 83 References Appendix: Conversion of Asbestos Mass Concentrations to Fiber Number Concentrations 0000JL UCC 000560 I. EXECUTIVE SUMMAHY Estimates of probable excess risks of lung cancer and meso thelioma are made for low levels of exposure to airborne asbestos under environmental conditions. This is done by extending to low doses the estimated risks observed for Industrial workers who were exposed at substantially higher doses. An extensive review of the dose-response data for industrially exposed workers (extending an earlier review by Nicholson) leads to the conclu sion that at an exposure of 1 fiber x years/ml the expected lifetime excess risk of lung cancer would lie in the following ranges: Smokers Nonsmokers r-i 1Ho o 1HO Men 10"3 to 10"3 Wanen 3 x 10_4,*to.3 x 10"3 4 x 10"5 to 4 x IQ"4 is some empirical evidence that these risks are not dependent on the age when exposure starts. For pleural and peritoneal mesotheliomas, the corresponding excess risk after half a lifetime's exposure (36-37 years) is estimated to be between 1 and 6 x 10"4. This risk appears to be independent of smoking habits (and not clearly dependent on sex). However, it is strongly dependent on the elapsed time since first exposure. 00003 1 UCC 000561 These estimates are derived from five independent studies which yielded reasonably consistent results. Two studies which yielded lower estimates of risk per unit exposure are discounted because of methodological deficiencies. Two other studies of workers exposed during mining and milling operations yielded much lower estimates of risk per unit exposure, perhaps indi cating lower biological activity of asbestos at early stages in processing. These risk estimates are extended to environmental exposure levels, using a number of assumptions about relationships between response, dose, age, and timing of exposure. There is consid erable empirical and theoretical evidence to support the assump tion of linear, nonthreshold relationships .between dose and. response, for both lung cancer and mesotheliomas. Data for mescthelianas fit reasonably well to a 'quadratic residence time" model, in which cumulative risks depend approximately on the third power of elapsed time since exposure. Response time relationships for lung cancers are less consistent, so that environmental risk estimates are more uncertain. Lifetime risk estimates for the general population are presented in Table 11 in Section V. For all categories except male cigarette smokers, lifetime risks for mesothelioma are predicted to be higher than those for lung cancer. For mesothe liomas, the most important period of exposure is expected to be the first 20 years of life, but most effects are expected after age 55. "Virtually safe" concentrations (i.e., exposure 00004 2 UCC 000562 levels corresponding to excess lifetime risks of 10") are pre dicted to be in the range 10" to IQ-4 fibers/ml for outdoor exposure, and 10" to 10" fibers/ml for indoor exposure. These concentrations are well below current detection limits. These estimates cannot be expressed reliably in units of ng/ml, becaus of wide variability in measured conversion factors. These estimates are subject to considerable uncertainties. The 10-fold range in the estimates that are presented reflects the variability in the results of occupational studies, and probably reflects the variability in the properties of asbestos. However, the estimates of risk may be much too high to apply to asbestos at an early stage in processing. They would also be much too high if the assumption of a linear, nonthreshold dose-response relationship proves incorrect. Risks of gastrointestinal and other cancers resulting from exposure to airborne asbestos have not been considered in detail in this report, but probably would add a small fraction (20-30%) to the estimates of excess risk that are presented. 00005 3 UCC 000563 II. INTRODUCTION This report has been prepared for the Institut ftir Wasser,3oden- und Lufthygiene des Bundesgesundheitsantes. It presents estimates of health risks, specifically risks of cancer, posed to the general population by low concentrations of asbestos present in the general environment. This report is part of the development of health criteria for airborne asbestos, on which ambient standards may be based. Accordingly, we have developed estimates of "virtually safe" airborne concentrations of asbestos. These are the concentrations of asbestos which, if inhaled throughout life, would be expected to give rise to an excess risk of cancer of about 10" (1 in a million). Sep arate estimates are provided for indoor and outdoor exposure, the sole difference being the difference in the length of time which the average person spends indoors and outdoors. The basis for the risk assessments presented in this report is the extensive information on excess frequency of cancers (primarily lung cancers and pleural and peritoneal mesotheliomas) observed in workers exposed to relatively high levels of asbestos in the workplace. These data have recently been reviewed by Nicholson (1981) , who derived a number of estimates of doseresponse relationships for lung cancer. In this report, we have extended Nicholson's review, primarily by deriving .-stimates of dose-response relationships for mesotheliomas, and by making 00006 5 UCC 000564 a preliminary analysis of relationships between response, age, and elapsed time. The results of these analyses are then used to derive esti mates of risks likely to be experienced by the general popula tion, exposed to much lower levels of asbestos from birth onwards. The extrapolation involves a number of assumptions about dose- time-response relationships. These assumptions are based on substantial empirical and theoretical evidence, but nevertheless introduce a number of uncertainties into the final estimates of general population risks. Uncertainties are also introduced by variability in the results of the epidemiological studies, which reflect variability in the properties of asbestos. The final estimates are presented to order of magnitude only. In preparing this report, we have greatly benefited from the pioneering efforts of Dr. William Nicholson in constructing dose-response relationships. We also acknowledge helpful advice from Dr. Nicholson and from Dr. Charles C. Brown of the National Cancer Institute. ' We emphasize that our analysis of dose-time-response rela tionships is preliminary, and that we have probably not extracted all the information that is available in published and unpublished studies. We recommend that a more formal and complete mathemat ical analysis of the available data should be conducted, to serve as the basis for improved estimates of general population risk. C0GC7 6 UCC 000565 III. BACKGROUND A. Nature of Asbestos Asbestos is the generic term for the fibrous forms of a group of naturally occurring mineral silicates of the amphibole or serpentine series. Asbestos occurs naturally in the seams / and veins of many igneous or metamorphic rocks, with fibers generally 1-20 millimeters in width and having length-to-width ratios of 100:1 and higher (NCI 1978, Advisory Committee Report 1979). Currently six fibrous silicates are generally classified as asbestos: the fibrous amphiboles--actinolite, amosite, anthophyllite, crocidolite, and tremolite; and the fibrous serpentine mineral--chrysotile ("white asbestos"). Chrysotile is the form of asbestos used most commercially. The chemical form of each type of asbestos is different as indicated below; Amphiboles Actinolite 2C aO.4M gO.PeO.8Si02.H20 Anthophyllite 7 MgO.8SiO 2.H20 Crocidolite Na20.Fe203.Fe0. 8Si02.H20 Tremolite Amosite 2CaO.5MgO.8Si02.H20 5.5FeO.1.5MgO.8Si02.H20 Serpentine Chrysotile 3Mg0.2Si02.2H2C SOURCE: Advisory Committee Report 1979, citing Hodgson 1965 00008 7 UCC 000566 Although each asbestos mineral is generally considered as a single chemical and physical entity, the fibers themselves belong to more complex crystal chemical systems, so that they may range in both chemical and physical properties (Langer 1974). Chrysotile fibers are generally of a smooth, silky texture and whitish in color; crocidolite fibers are straighter, shorter, and less silky than chrysotile and bluish in color; amosite fibers are more brittle than either crocidolite or chrysotile and brownish in color. It should be noted, however, that color is not a reliable indicator for the identification of asbestos type in any particular product. These minerals may all occur in the nonfibrous form and in such cases are not considered as asbestos. ... Asbestos fibers are flexible, possess high tensile strength, and are easily separated into filamentous strands. The proper ties which give asbestos its major commercial value are insolu bility, noncombustibility, strength, ability to be woven, and efficacy as a binding or reinforcing agent when combined with other materials such as plastic or cement. In varying degrees, the different types of asbestos are also efficient at absorbing sound and resistant to electric currents, high temperatures, and alkalis. In addition, the amphibole fibers are resistant to acids. B. Uses Asbestos fibers can be adapted to a large number of uses due to their high tensile strength, flexibility, chemical and CISCO? 8 UCC 000567 heat resistance, and favorable frictional properties. Dependent upon the properties of the fibers, asbestos can be spun, carded or woven; used for structural reinforcement of materials such as cement, plastic, and asphalt; or laid and pressed to form paper. The Asbestos Information Association suggests that there are more than 3,000 uses. Longer asbestos fibers tend to be used in textiles, electri cal insulation, filtration media, and maximum-strength asbestos products. Medium length fibers are used as reinforcing fillers in asbestos cement products, in friction materials such as brake linings, in pipe covering, and in paper. Shorter fibers tend to be used as reinforcing fibers in plastics, asphalt, and floor tiles, as well as in oil-well drilling muds and paints (NCI 1978) . C. Occurrence . Asbestos formation occurs as a result of several stages of geologic processes in which parent rocks (ultramafic, dolomi- tic, or limestone) are transformed into the substance. Transfor mation of parent rock may occur under localized conditions of temperature and pressure or as a result of the action of hot mineral solutions that can alter some minerals to form others. In addition to commercial mining and milling, the occurrence of asbestos in the environment may result from the inadvertent disturbance of asbestos deposits by activities as road building or farming; from the manufacture, use, repair and demolition of asbestos-containing products; from the transportation of COCIO 9 UCC 000568 asbestos ore, milled fiber, products, and wastes; and from the disposal of asbestos waste. Individuals can be occupationally exposed to the substance during processing, including mining, milling, factory production, and handling of asbestos-containing products. Contamination of air with asbestos can result from the spraying of the material for fire protection or acoustical purposes. Workers may also be exposed during end-product use such as brake maintenance or application of insulation. Other examples of workers poten tially exposed to asbestos include those in ship-building, gen eral construction, demolition, building inspection, repair and maintenance (of asbestos-containing structures), waste disposal, and cleaning operations. Workers' family members may be exposed to asbestos from dusts brought home on clothing, shoes and hair. People can be exposed in the home as a result of performing construction and repairs using asbestos containing materials. Such materials include asbestos plastic-fillers, cement sheets, and furnace and heating equipment cements. . Finally, asbestos can be present in the general environment in air, water, food, and beverages. Asbestos concentrations in the range of 10-100 ng/m have been reported in the general urban atmosphere. Near some factories using asbestos, asbestos concentrations have been reported to be as high as 5,000 ng/m^. The air inside buildings can be contaminated due to working with or damage or erosion of asbestos material. In addition, asbestos has been detected in some samples of drinking water, eocii 10 UCC 000569 food, beverages, and pharmaceutical preparations (NCI 1978, Advisory Committee Report 1979}, D. Routes of Exposure The most important route of exposure of asbestos fibers is by inhalation into the respiratory tract. The disposition of fibers entering the respiratory tract is not fully understood. Some fibers can be conveyed to the gastrointestinal tract by airway clearance mechanisms and others to the pleural and peri toneal cavities by lymphatic drainage. The physical character istics of asbestos fibers which favor penetration of asbestos fibers to the bronchi, lungs, and pleura are briefly discussed in Section III.F below. Ingestion is another potential route of exposure. It may be secondary to inhalation of fibers wherein the individual swallows nasal and bronchial secretions containing inhaled as bestos fibers. It is assumed that this is the form of inges tion which accounts for the excesses of gastrointestinal cancers which have been observed in some groups of asbestos workers. Ingestion of asbestos can also occur by the consumption of con taminated water, food, or beverages. It should be noted, how ever, that most asbestos fibers entering the gastrointestinal tract are probably excreted with the feces. Asbestos may be injected inadvertently through the use of asbestos-containing syringes or through the use of fluids containing pharmaceutical agents which were filtered through asbestos pads during manufacture. After injection into the 00C12 UCC 000570 bloodstream, asbestos is rapidly removed and deposited into various tissues, with the highest concentrations observed in lungs, liver and spleen (NCI 1978). E. Problems in Identification of Asbestos Fibers The identification of asbestos fibers requires structural, morphological, and chemical information. Due to the variability in the characteristic properties of asbestos fibers, their de tection and identification can often be difficult. Several methods have been proposed for the identification and quantita tion of asbestos in air, water, and biologic materials. These include optical and electron microscopy, X-ray diffraction, differential thermal analysis, and infrared, emission and atomic absorption spectroscopy. A number of factors contribute to the difficulty in the analysis of asbestiform materials. The problems which can be encountered in such analyses include: complexity of sample preparation of the differing -media; level of fiber exposure (contamination); presence (kind and amount) of other particles; variations in reactions of tissue types to different preparation techniques; improper choice of instru mental technique; and the great amount of time required for analysis and quantitation (Langer 1974). F. Identification of Biologically Active Fractions The information available about the disposition of asbestos fibers following inhalation or ingestion is relatively limited due to the difficulties involved in the assay of biological C(ri3 12 UCC 000571 tissues for the substance. The factors which have been consid ered to influence fiber deposition and the occurrence of human disease include fiber length, fiber diameter, fiber number, stability of fiber in vivo, fiber translocation and migration, fiber type, and overall dose. | It has been suggested that fibers less than 5 microns in length may be completely phagocytosed in vivo whereas those longer than 25 microns generally are not (Allison 1973). Those fibers in the 20 micron range may be only partially phagocytosed or may cause thinning of the phagosomal membrane (Langer et al. 1979). The predominant fibers found in the lung parenchyma and in extrapulmonary organs are those shorter than 5 microns in length. However, in cases of asbestosis, peribronchial and perivascular lesions generally contain both long and short fibers. The interaction of short and long fibers has also been considered as an aspect of biological activity. In summary, current data suggest that both long and short fibers are biologically active and suspect in producing human disease. The diameter of asbestos fibers appears to play an important part in determining biological activity. The diameters consid ered most important range from 0.5-2.5 microns (Langer et al. 1974) When compared with "thin" fibers, "thick" (and long) fibers tend to produce less biological response. Reduction in fiber diamete*. tends to reflect an increase in particle number and surface area per unit mass of material. This suggests that when evaluating environmental samples, any diameter under several Wll UCC 000572 microns should be considered in the active respirable range (Langer et al, 1979). An increase in asbestos fiber number may in some cases increase and in others greatly decrease biological activity. A correlation of activity and absolute number is probably not equal for all fiber types. Furthermore, such a correlation is probably not equal for all disease entities. It has been proposed that fiber retention may be partially related to fiber stability in vivo. Resistance to degradation has been considered to be of prime importance in the induction of cancer (Pott 1978). However, fiber degradation and dissolu tion in vivo may also increase the activity of the fiber (Langer et al. 1979) . Approximately 10-30% of asbestos fibers retained by human lungs become coated with hemosiderin and mucopolysaccaride to form yellow-to-brown rod-shaped structures called asbestos bodies (NCI 1978). The effects of asbestos fibers and bodies are not limited to one target organ? they have been found in almost every extrapuLmonary tissue. In certain cases removal of fibers from the lung may result in translocation to another site. Short fibers tend to be translocated more readily than long fibers and should therefore be expected to lodge in organs other than the lungs. It has been difficult to assign a scale of relative pathoge nicity to various asbestos types. This is related to many fac tors, including differences in physical characteristics within C0C15 14 UCC 000573 fibers of a particular asbestos type; changes in fiber charac teristics during processing; and contamination of one type of asbestos with another type. It has been suggested that chrysotile is less hazardous than other asbestos types. However, high rates of lung cancer in asbestos workers have been related to all types of asbestos (see Section IV below). In addition, pleural and peritoneal mesotheliomas have been observed in workers exposed primarily to chrysotile, as well as to crocidolite and amosi te. G. Health Effects The following discussions of carcinogenic and noncarcino- genic health effects are based upon data presented in the fol lowing reviews on asbestos: NCI 1978, NIOSH 1976, and Advisory Committee Report 1979. 1. Carcinogenic effects--human studies , There is extensive evidence that asbestos is a cause of lung cancer in humans. The majority of this evidence is derived from occupational epidemiological studies. It should be noted, however, that the lung tumors which occur in persons exposed to asbestos have no special diagnostic feature. It is usually impossible, therefore, to determine to what extent any specific lung tumor is related to asbestos, rather than to other agents such as cigarette smoke or polluted air to which the individual may have been exposed. The combined effect of smoking and as bestos exposure has been found to produce a risk of lung cancer that exceeds the sum of their separate risks. 15 UCC 000574 Asbestos exposure has also been strongly correlated with pleural and peritoneal mesothelioma. The disease is one of the rarest malignancies and occurs mainly in workers in heavy industry, as in shipbuilding. Mesothelioma rarely occurs in the absence of exposure to asbestos dust. It is therefore much easier to distinguish its relationship to asbestos than is the case with lung cancer, due to the strong association of the latter with tobacco smoking. Several epidemiological studies have suggested that an excess risk of cancer of the digestive system is associated with occupational asbestos exposure. A problem with some of these studies has been sketchy clinical and pathological data and the possible inclusion of peritoneal mesothelioma cases among all observed cases. This can make it difficult to document .an increased risk of any one digestive system cancer independent of that for mesothelioma. It must be stated, however, that the excesses are larger than can be reasonably explained away in this manner, so that a true increase in gastrointestinal cancer seems more probable. Highly suggestive but limited evidence exists suggesting a casual link between asbestos and laryngeal cancer. The risk, in absolute terms, however, is much smaller than for the cancers already described. There are currently insufficient data avail able to permit adequate study of possible synergism between the effect of smoking and asbestos exposure on laryngeal cancer incidence. 00C17 UCC 000575 2. Noncarcinoqenic effects of asbestos Asbestosis is a disease characterized by a diffuse inters titial fibrosis of the lung, which may or may not be associated with fibrosis of the outer or inner layer of the pleura. By definition, the disease is specifically related to asbestos exposure. The disease is difficult to diagnose because its onset is gradual, and the characteristic signs and symptoms may occur in other lung diseases and are somtimes difficult to detect. Symptoms rarely become apparent until after at least a decade of exposure. By the time the initial symptom of dyspnea on exertion appears, the underlying disease process is well established. At present, there appears to be no scientific evidence that asbestosis and lung cancer are interrelated, except that they are both causally associated with asbestos exposure. Benign pleural effusion (fluid development in the space between the pleural membranes lining the lung and chest wall) is another manifestation of disease associated with asbestos exposure. It may occur in the presence of sane degree of paren chymal asbestosis or with other asbestos-related lung diseases or as the only and most prominent process. Pleural plaques and calcification (thickening) are additional manifestations of asbestos exposure, which may occur alone or in association with fibrosis of the lung. The plaques do not generally alter pulmonary function to a significant extent; pleural thickening may be symptomless or associated with restriction of breathing and chest discomfort. 00C18 UCC 000576 IV. DOSE-RESPONSE DATA DERIVED FROM STUDIES OF INDUSTRIAL WORKERS A. Introduction Ideally, one would like to base estimates of risks to a human population on dose-response data derived from studies in the same population. Unfortunately, this can rarely be done. The general population is usually exposed at very low doses, so that effects are likely to be small and hard to discern. Not only are members of the general population exposed to low doses of the material of concern, they are also likely to be exposed to many other things, usually also at low doses--which will also have some small effects--making the sorting out of the different effects nearly impossible. Because of these prob lems of low doses and of confounding with other exposures it becomes necessary to turn to populations that are more heavily exposed--usually worker populations, exposed on the job. Worker populations often show clear-cut effects at the high doses to which they are exposed. From the dose-response curves observed it is then necessary to infer dose-response curves that are likely to be valid at the low levels of ambient exposure. There are difficulties with making such inferences. Industrial workers are usually men. They do not include the very young or the very old. They are usually healthier than the general population. They are often exposed to more than one toxic material in their working or personal lives. For example, many are cigarette smokers. Thus there are logical C0C19 19 UCC 000577 and conceptual as well as biological problems in extrapolating data from young, healthy men to other categories of people,' such as children, pregnant women, or older, not nearly so healthy retired persons. Within these limitations, however, it is possible to make seme reasonable use of the data that have been developed on industrial workers. Such data are useful when one can find workers exposed to clearly different levels of the material. Short-term exposures have occurred. High exposures lead to earlier illness and earlier recognition of health effects. Workers, at least male workers, are relatively easy to trace. However, in addition to the problems pointed out earlier, the major gap in the worker-derived data is that they do not give any direct measures of effects of exposures at young ages, espe cially under age 16. Rarely do industrial exposures begin before age 16. Thus there are assumptions that need to be made and some uncertainties associated with them. Nicholson (1981) has brought together and reviewed all the published industrial epidemiology studies on workers exposed to asbestos. His review provides a basis for the analysis pre sented here. In the material that follows an attempt has been made to state all the assumptions that enter into the computa tions and, where important, to indicate the variation that would ensue from the use of other possible assumptions. C0C20 20 UCC 000578 B. Measurement of Health Effects This report deals largely, but not exclusively, with data on mortality--death from specific causes. This leads to some immediate problems. The major causes of (cancer) deaths asso ciated with asbestos exposure have been cancer of the bronchus and lung (ICD 162), mesotheliomas, both pleural and peritoneal, and cancers of the esophagus (ICD 150), stomach (ICD 151), colon (ICD 153) , and rectum (ICD 154). Some increases have also been reported in the frequency of cancers of the larynx (ICD 161), pharynx and buccal cavity (ICD 140-149) , and cancer of the kidney (ICD 130). Selikoff et al. (1979) have reported that for all other cancers the ratio of observed to expected deaths, as reported on death certificates, was 1.91--a significant excess. - 1. Ascertainment of cause of death Two difficulties arise in using data on these causes of * death. The first is that reports on death certificates can be substantially in error. In a detailed review of death certifi cate reporting compared to what they called "best estimates" from detailed hospital data, Selikoff et al. (1979) found that the death certificates substantially over-reported cancers of the pancreas (ICD 157) , liver and biliary passages (ICD 155), and possibly brain (ICD 193). There was substantial under-reporting of mesotheliomas, and about a 10% under-reporting of cancer of the lung. Mesotheliomas, being rare causes of death, are usually not reported in national death statistics. /"? ft J 1 'J f " -M** 21 UCC 000579 There have been reporting problems for mesotheliomas in the past. These are likely to continue in the future. Unless specified as "malignant," mesochelicmas were usually coded to ICD 228, "benign" tumors. In addition, most national data are not published in more detail than the basic 3-digit codes. Thus, data on mesotheliomas of the peritoneum, ICD 158.9, were not published in the United States. Similarly, data on mesothe liomas of the pleura were not published and, under certain cir cumstances, all mesotheliomas not otherwise specified were in cluded under code 199 "malignant neoplasm without specification of site." Special tabulation of cause of death by 4-digit code can usually be made by national vital statistics centers. When comparisons have been made of death certificate report ing and hospital diagnosis in the United States, lung cancers have been found generally to be properly reported (see Percy et al. 1981). In one study of about 50,000 cancer deaths, hos pital records recorded 10,059 lung cancer deaths, and the death certificates recorded 10,178 lung cancer deaths. Several other sites, however, were not as well reported (see Table 1). The second important point to recognize is that there appear to be a strong interaction with cigarette smoking in the develop ment of lung cancer in asbestos-exposed workers. In typical circumstances, the frequency of lung cancer in cigarette smokers is about 10 times that in nonsmokers with similar exposure 0002? 22 UCC 000580 TABLE 1 COMPARISON OF REPORTED FREQUENCIES OF CERTAIN TYPES OF CANCER FROM HOSPITAL RECORDS AND DEATH CERTIFICATES Site (and ICD number) Hospital Records Death Certificates Percent Difference Buccal cavity (140-149) Colon (153) Rectum (154) Liver (155.0, 197.8) Kidney (189) Brain (191, 192) 1,397 4, 546 2,098 462 984 1,074 1,187 5,131 1,36 7 687 930 1,171 - -15 +13 -35 +4 9 -5 +9 SOURCE: Percy et al. (1981) to asbestos (Selikoff et al. 1979). For this reason, measures of excess risk of lung cancer in asbestos workers.are likely to be variable unless account can be taken of the prevalence of smoking among them. No interaction with smoking has been reported for mesotheliomas, and as shown below, estimates of risks for mesotheliomas appear to be less variable than those for lung cancers. 2. Dependence of disease frequency on age and elapsed time since first exposure Cancer is clearly a disease of older ages, with roughly half of all cancer deaths occurring ir persons over 65. It is also a disease of long delay; there is a long time from ini tial exposure in a working population to the diagnosis of cancer, 00023 23 UCC 000581 or to death from cancer. (Median time from diagnosis to death is about 2 years, for the totality of cancers.) Selikoff et al. (1979) have reported on deaths from lung cancer and mesotheliomas among insulation workers in the United States and Canada as a function of the number of years following initial exposure. For lung cancers, the median time to death (from first exposure) was about 32 years. For mesotheliomas, the median time was sanewhat longer--36 years. It has been suggested that the effect of the interaction of cigarette smoking with asbestos exposure is to move forward in time the appearance of the lung cancers. Under this hypothesis, nonsmoking asbestos workers might eventually develop as much cancer as the smokers, but, on the average, the disease might appear several years later. Mesotheliomas, which appear to be unaffected by cigarette smoke, lead to later deaths than do the lung cancers. . Seidman et al. (1979) examined the effects of the age at which the men started working in an amosite factory on their subsequent development of lung cancer. Their conclusion was that ...heavy direct exposure to men who are already at "cancer ages" [i.e., men 50 and over] can result in a very large increase in mortality in a surpris ' ingly short period of time, 5-14 years after onset of work. With lighter direct exposures, even to men already at "cancer ages" there is a latency period of a number of years before significant increases in mortality are seen and these increases are smaller than those of the more heavily dosed men. Regardless of the level of exposure, there were few deaths in younger men, those in their 20's at the 5 year point after onset [of exposure], 0.0C24 24 UCC 000582 The length of the latency period depends not only on the dosage but also on the age at which expo sure takes place. This latter observation is consistent with the hypothesis that asbestos may act at a middle or late stage in the develop ment of cancer/ leading to the clinical development of cancers initiated earlier in life. The manner in which asbestos leads to cancer has also been considered by Day and Brown (1980). Analyzing the data of Seidman et al. referred to above, they came to the conclusion that as bestos operates as an early-stage, rather than a late-stage carcinogen for workers employed for less than 2 years. However, they qualified this conclusion by pointing out the unusual manner of "residence" of asbestos. ...there is some ambiguity in interpreting the evolu tion of risk after the external exposure has stopped, because asbestos bodies remain in the lungs. It might be more appropriate to consider exposure as being continuous but at a decreasing level as the asbestos bodies are either excreted or their carcino genic ability is reduced. However, the mechanism of action of asbestos carcinogenesis is unknown, and it is not clear that its carcinogenic effect remains with the asbestos bodies that remain in the lungs. The observed increasing excess cancer inci dence is thus consistent either with an early-stage effect or with a late-stage effect resulting from the asbestos remaining in the lungs. Seidman et al. did not analyze the age-at-first-exposure effect for mesotheliomas, but Day and Brown, examining other data, concluded ...the quadratic residence time model as described by Peto appears to fit the observed data reasonably well. Under this model, the incidence at time T, e.g,, I (T) , is proportional tos/pT (T-t) ^c (t} dt, where c(t) is the exposure level at time t. It should 00025 25 UCC 000583 be noted that this quadratic residence time model is mathematically equivalent to a model derived from the assumption of a first-stage effect in a four- stage carcinogenic process. As with the risk for lung cancer following asbestos exposure, we can inter pret these data as describing either the effect of a continuous exposure due to the inhaled fibers or the effect of short-term exposure to an early-stage carcinogen. : We shall deal with the quadratic residence time model later, when we attempt to estimate lifetime risks of mesotheliomas following exposure to asbestos. Thus, while it is clear that age at first exposure is of considerable consequence, few of the industrial data reported assemble the information in this manner. For problems of ambient exposure, it is most important to attempt to make estimates of the effects of life-time expo sures, beginning at birth, if that is when first exposure occurs. 3. Measures of excess risk There are three questions in the measurement of excess risk that must be considered. a. Excess over what? What comparison population is appro priate? If we deal with exposed workers, should they be compared with other unexposed workers, or the general pop ulation, or people in the same region or city, or same ideal, low-risk population? Usually comparisons are made with a general, country wide population, at least as a first step. This usually leads to an understatement of risks for a working popula tion, because working populations are often healthier than nonworking populations. When deaths in workers are compared 00026 26 UCC 000584 to expected deaths in the total population, one often finds that the workers show about 80% of the expected deaths computed from general population rates for persons of the same age and sex, b. Excess among all causes Of death, or only certain causes of death? If earlier work or if animal studies have shown excess deaths from certain specific causes, should these be the only ones considered? Are excess deaths from one cause made up for by reduced deaths from other causes? It is most common to look both at total causes of death and at specific causes that are likely to be asso ciated with the exposure. Biological plausibility is often invoked to justify examining one cause and not another. Thus, in asbestos exposure, it looks most reasonable to look at respiratory disease and probably noninfectious diseases of the digestive tract. It would be suprising to find increased leukemias, or increased diabetes, but not surprising to find increased pneumonias or bronchitis, or even cancers of the colon and rectum. c. What specific measures of excess should be used? Relative risks (or Standard Mortality Ratios, SMR) or absolute increases? Average years of life lost because of premature death? The relative risk or Standard Mortality Ratio is a commonly used measure. It compares observed deaths with expected deaths, where "expected" deaths are usually calculated 00C27 27 UCC 000585 from general population data. Thus for the average ciga rette smoker (about 20 cigarettes per day), the relative risk of death from lung cancer is about 10 or 11 compared to a nonsmoker. (The SMR would then be 1,000 or 1,100. A SMR of 100 means observed equals expected.) The SMR is a useful measure for excesses among total deaths and for relatively common causes of death such as lung cancer. For causes of death that are rare in the general population such as mesotheliomas, the SMRs would approach infinity and would convey little information. For mesotheliomas, therefore, absolute increases in death rates due to this cause are used as the appropriate'measure. 3ecause meso theliomas appear to be a disease that appears later in life, on the average, than lung cancer, it is important * that the ages of the exposed populations be well defined, so that various populations may be compared on an equal basis. 4. Biases and confounding factors There are four major possibilities for distorting estimates of excess risk through problems of confounding (with other expo sures) or statistical bias through the use of unrepresentative samples of the exposed or comparison populations. a. "Healthy worker effect." This effect has already been mentioned. Several suggestions have been made to minimize or avoid this effect, but none seems fully satisfactory. These suggestions range from arguing that no adjustment 00028 28 UCC 000586 need be made, to suggestions that all individual SMRs be inflated by the inverse of the 5MR for all causes, for example, by multiplying cause-specific SMRs, such as that for lung cancer, by 1/SMR for the total deaths. Thus, if the SMR for total deaths were 0.8 and the SMR for lung cancer was 1.2, the "adjusted" lung cancer SMR would be 1.2 x 1/0.8 * 1.50. Another suggestion is that the SMR for a cause of death thought very unlikely to be related to exposure {such as diabetes) be used as a "normalizing" divider for the cause-specific SMR. In the example above, say the SMR for diabetics was 0.75. The adjusted SMR for lung cancer would be 1.2 x 1/0.75 * 1.6. b. Other characteristics of the workers. Most workers are urban dwellers. Urban residents have, in general, higher cancer rates than rural persons. The.differential increases with the density of urban populations, although there is no reason to believe that population density, in itself, should be cancer-causing. Similarly, social class has been associated with increased cancer rates. Social class, however, is difficult to disentangle from occupation, since occupation often determines social class. Long-time patterns of diet, recreational activity (or lack of it) and other disease-related factors are associated with Bocial class. In general, when generalizing from industrially exposed populations to more generally exposed populations, issues of urbanization, population density, - 29 UCC 000587 social class, etc., are usually set aside with the assump tion that the industrial worker is nearly representative of the general population with respect to these issues. c. Cigarette smoking. This is particularly important with respect to asbestos-related disease. There is some evidence of excesses in cigarette smoking among asbestos workers compared to the general population (Hammond et al. 1379, Seidman et al. 1979). Patterns of cigarette smoking in FDR appear to be similar to those of the United States (Erben 1979, Tobacco International 1980) so that computations based on American experience should be appli cable to FDR. Overall, in 1979 in FDR approximately 41% of men and 34% of women were reported as cigarette smokers. (Tobacco International reported higher rates in men and lower rates in women.) The proportion of adults smoking cigarettes fell off with increasing age, as in the United States. In FDR, the age group with the highest proportion of persons smoking was under 30 and the lowest proportion was in persons over 60. In part the low proportion of smokers in persons over 60 is due to the higher death rates in smokers, particularly among the older age groups. d. Competing causes of death. The smoking effect--particularly in older persons--leads to deaths from causes that may not usually be associated with asbestos exposure. This has two effects on the data that are examined for asbestos effects. First, there is an artificial increase 00030 30 UCC 000588 in so-called "natural" {unassociated) deaths, and a conse quent underestimate of the associated deaths. Second, there is evidence of a "saturation" effect with increasing dose. For example, after seme effective dose level, further exposure may not lead to deaths from lung cancer, but may lead to earlier deaths from asbestosis. For this reason, the dose-response curve would flatten or even turn down at very high doses, because persons died early of other causes. This saturation effect, due to competing causes of death, is also seen when followup proceeds for a long time. Nicholson (1981) , in considering relative risks for asbestos workers for all causes of death, found the maximum relative risk after about 25-30 years of followup and then a slow decline in relative risk with continuing followup. Thus, a followup period of 25-30 years appeared to provide the best measure of the excess risks of lung cancer. Shorter followup periods yield underestimates of risk because the latency period has not been reached; longer followup periods also yield underestimates of risk because of competing causes of death. The latter phenomenon has not been discussed extensively in the epidemiological literature. 5. Conversion of risk measures to estimates of lifetime risks Epidemiologic studies rarely measure lifetime risks. To do so would require following a population to extinction which. 00031 UCC 000589 of course, is rarely done. To estimate lifetime risks, two principal approaches can be used. a. The first approach is useful when the agent increases the frequency of a disease which is fairly frequent in the gen eral population. Then, if the relative risk R (frequency in exposed population/frequency in general population) can be as sumed to be constant (independent of age), the lifetime risk in the exposed population can be estimated from the lifetime frequency of the disease in the general population. Specific ally, if the lifetime probability of developing the disease in the general population is pB - i - then the lifetime probability of developing the disease in the exposed population is PE - 1 - e'8i. For low values of R, P = RPg. This method of estimating life time risk is appropriate for asbestos-induced lung cancer, for which there is empirical evidence that R is approximately con stant (see Section IV. below). b. In cases in which the relative risk varies with age, or cannot be calculated because the disease is rare in the gen eral population, lifetime risks can only be calculated by assuming a specific model for the dependence of risk on age and duration of exposure. For many cancers, there is empirical evidence that frequency depends on the k'th power of age (and/or elapsed time since exposure), where k is usually between 3 and 6 (Armitage 00033 32 UCC 000590 and Doll 1961). This is the basis for a number of models from which lifetime risks can be calculated (Armitage and Doll 1961, Day and Brown 1980). This approach is necessary for asbestosinduced mesotheliomas, which are rare in unexposed individuals. A general problem with the approach is that the results are strongly dependent on the value assumed for k, because most of the lifetime risk falls in the last quarter of life for which direct observations are often scanty. However, this problem is not very serious for asbestos-induced mesotheliomas, for which there are good data on frequency up to 35-40 years after exposure, and for which k can be estimated empirically {Peto 1979, Day and Brown 1980) . Peto (1979) has applied the first approach to asbestosinduced lung cancer and the second to asbestos-induced mesothe liomas. For mesotheliomas, Peto utilized a "quadratic residence time" model (see footnote to Table 3); this is equivalent to assuming k * 3. Tables 2 and 3 present the results of Peto's calculations, modified to present lifetime risks as a function of age at first exposure and duration of exposure. (Note that the calculations are designed only to illustrate the functional form of these relationships; the exposure level required to cause this response is hypothetical.) In the case of lung cancer (Table 2) , the calculations assume a linear dependence of lifetime risk on duration of expo sure (i.e., on cumulative dose), but predict that lifetime risks are independent of age at first exposure. In the case of 0.0033 33 UCC 000591 mesotheliomas (Table 3), however, the calculations predict that lifetime risk depends very strongly on the age at first exposure, and that only the first 20 years of exposure contribute significantly to the lifetime risk. The models on which these results are based are equivalent to assuming that asbestos acts at a middle or late stage in the development of lung cancer, but at an early stage in the development of mesothelioma. As discussed elsewhere in this report, there is substantial empirical basis for the functional forms of the dependence of risk on age and duration of exposure (Nicholson 1981), but there is ambiguity about the stages at which asbestos acts (Day and Brown 1980) . Accordingly, although we will use assumptions equivalent to Peto's in calculating lifetime risks, it should be recognized that these assumptions may exaggerate the differences in age-dependent frequency between lung cancers and mesotheliomas. An important result in Table 3 is that the lifetime risk for mesotheliomas is 2-7 times higher in persons exposed from birth than in persons starting exposure at ages 15-35. This is important because all the data used in Section IV are derived from workers whose exposure started after age 16. Hence a multiplying factor is needed when these data are used in Section V to estimate risks to persons in the general population, whose exposure to ambient asbestos starts at birth. The figures in Table 2 would indicate that no such multiplying factor is needed for lung cancer, but this conclusion would be wrong if asbestos actually acts /it an early stage in development of lung cancer. '` 34 OOC34 UCC 000592 TABLE 2 THEORETICAL DEPENDENCE OF LIFETIME RISK (%) OF LUNG CANCER ON DURATION OF ASBESTOS EXPOSURE AND AGE AT FIRST EXPOSURE Duration Of Exposure {yr) 0 15 Age at First Exposure 25 35 45 55 1 0.04 2 0.06 5 0.19 10 0.38 20 0.76 3 0 1.1 40 . 1.5 50 . 1.9 60 2.2 0.04 0.08 0.19 0.38 0.76 1.1 1.5 1.8 0.04 0.08 0.19 0.39 0. 77 1.1 1.4 . 0.04 0.08 0.20 0.39 0.76 1.1 0.04 0.08 0.20 0.38 0.69 * 0.04 0.08 0.18 0.33 SOURCE : Modified from Peto (1979) Assumptions: relative risk independent of age, but propor-tional to exposure level x duration; lifetime 73 yrs. Peto presented these risks as hypothetically resulting from continuous exposure to 2 fibers/ml. 00035 35 UCC 000593 TABLE 3 THEORETICAL DEPENDENCE OF LIFETIME RISK {%) OF MESOTHELIOMA ON DURATION OF ASBESTOS EXPOSURE AND AGE AT FIRST EXPOSURE Duration of Exposure (yr) 0 15 Age at First Exposure 25 35 45 55 1 0.42 0.21 0.12 0.06 0.03 0.01 2 0.82 0. 41 0.23 0.12 0.05 0.02 5 1.90 0.93 0.52 0.26 0.11 0.03 10 3.21 1.6 0.89 0.43 0.17 0.05 20 5.0 2.5 1.3 0.59 0.21 30 5.8 2.9 1.5 0.63 40 6.0 3.0 1.5 50 6.0 3.0 60 6.0 SOURCE: modified from Peto (1979) Assumptions: contribution of exposure at time t to risk at time T is proportional to (T-t)1 ("quadratic residence time model"); lifetime * 73 yrs. Peto presented these risks as hypothetically resulting from continuous exposure to 2 fibers/ml. 00036 36 UCC 000594 C. Measurement of exposure Problems in measuring exposure to asbestos have been reviewed recently by Nicholson (1931), whose comments are reproduced here in their entirety. 1. Importance of past estimates The accuracy' of judgements on dose-response relation ships for asbestos is largely limited by our knowledge of past fiber exposures of those populations whose mortality or morbidity were later evaluated. Few measurements were made in facilities using asbestos fibers prior to 1970. Further, those measurements that were done usually quanti fied all dust (both fibers and particles) present in the workplace air. Current techniques, using membrane filters and phase contrast microscopy for the enumeration of fibers longer than 5 micrometers, have been utilized in Great Britain and the United States only since 1964 (Holmes, 1965; Ayer et al., 1965) and have been standardized in the United States only since 1972 [National Institute for Occupational Safety and Health (NIOSH) , 1972 ; 1979], These techniques may be utilized to evaluate work practices and conditions believed to be typical of earlier activities. However, it is always difficult to duplicate materials and conditions of earlier decades and such retro spective estimates are necessarily uncertain. Alternatively, fiber counting techniques and the instrumentation of earlier years can be used together to simultaneously evaluate a 00037 37 UCC 000595 variety of asbestos-containing aerosols. The comparative readings then serve as a "calibration" of the historic instrument in terms of fiber concentrations. Unfortunately, the calibration depends on the type and size distribution of the asbestos used in the process under evaluation and the quantity of other dust present in the aerosol. Thus, no universal conversion has been found between earlier dust measurements and current fiber counts. In the United States and Canada those few data that were obtained on asbestos workers' exposures prior to 1965 were based upon total dust concentrations using a midget impinger. Fibers were inefficiently counted with this instrument because of the use of bright field microscopy. Attempts to compare fiber concentrations with midget impinger particle counts generally showed poor correlations (Ayer et al., 1965; Gibbs and Lachance, 1974). In the United Kingdom the thermal precipitator was used from 1951 through 1964 in one plant for which environmental data have been published. This instrument, too, does not allow accurate evaluation of fiber concentrations and the variability in the correlation between fiber measurements and thermal precipitator data is reported to be large (Sykes, 1977, quoted in Advisory Committee, 1979b), but no specific data are available. ' ' 00C38 38 UCC 000596 2. Measurement techniques Even with the advances in fiber counting techniques, significant errors may be introduced into attempts to for mulate general fiber dose-response relationships. The convention now in use, that only fibers longer than 5 micro meters be counted, was chosen solely for the convenience of optical microscopic evaluation (since surveillance agen cies have such instrumentation) . It does not necessarily correspond to any parameter of biological importance for asbestosis, lung cancer, or mesothelioma. While it is readily understood that counting only fibers longer than 5 micrometers enumerates but a fraction of the total number of fibers present, there is incomplete awareness that the fraction counted is highly variable. It depends upon the fiber type, the process or products used, and even the past history of the asbestos material (old vs. new insula tion material, e.g.), among other factors. For example, the fraction of chrysotile fibers longer than 5 micrometers in an aerosol can vary by a factor of ten (from as little as 0.5% of the total number to as many as approximately 5%). When amosite aerosols are counted, the fraction longer than 5 micrometers may be 301, extending the variability of the fraction counted to two orders of magnitude (Nicholson et al., 1972 ; Nicholson, 1976 ; Winer and Cossette, 1979). Thus, even perfect measurement of workplace air, with accu rate enumeration of fibers according to accepted methods, gcooo 39 UCC 000597 may introduce a significant uncertainty in the dose-response relationships for any specific asbestos disease when dif ferent work environments are studied. Those uncertainties that exist in the physical deter minations of past fiber concentrations and our difficulty in evaluating the exposure parameter of importance in cur rent measurements are exacerbated by the sampling limita tions in determining individual or even average exposures of working populations; only few workmen at a worksite are monitored and then only occasionally. Variability in work activities, in personal habits, and in sampling circumstances add considerable uncertainty to our knowledge of dose. This gloomy picture is not painted to suggest that one should throw up one's hands and say dose-response rela tionships are impossible to obtain or to require that any dose-response data be restricted only to those circumstances in which they were obtained. Rather, it is described to indicate how approximate any estimated dose-response rela tionship is likely to be and to emphasize that projections of health effects at low exposures, made on the basis of available information, may seriously mis-state these effects. D. Relating dose to response The usual procedure is to calculate cumulative dose in fibers/ml x years and then calculate response per unit of accumu lated dose. This model of disease can be written as; 00040 40 UCC 000598 Excess risk in period (t2 - t1) constant ft) dt t where c (t) is a function of the dose at time t. The model makes two important assumptions, which may or may not be true: 1. The dose-response relationship is linear. As dose increases, response increases, proportionately. 2. There is no dependence on the timing of exposure--that is, a fiber makes the same additive contribution to risk no matter whether it is inhaled early or late in life. Several sets of data seem to support the assumption of linearity (see Nicholson 1981, and further discussion in Section IV.E below). However, some dose-response curves appear to flat ten out at very high levels of exposure and at very long periods of followup. This may be due to dose saturation effects, to ccmpeting causes of mortality, or to both. The second assumption is less plausible, for several reasons: 1. Cancer is known to have a latent period after exposure before clinical disease appears, so that exposures late in life make little or no contribution to risk. 2. Asbestos fibers are retained in the body for long periods, so that a fiber inhaled early in life has a much greater opportunity to cause biological effects than one inhaled late in life. 3. The multistage model of cancer provides theoretical support for the assumption that cancer risks depend strongly on age and on elapsed time since exposure (Armitage and Doll 1961, Peto 1979, Day and Brown 1980; see discussion above). These considerations support several alternative formula tions of the mathematical model given above: 00041 UCC 000599 (i) to take account of a latent period, L, XExcess risk a c {t)dt; (ii) to take into account retention of fibers, Excess risk a or if fibers are cleared at a constant rate a, Excess risk a it c(T)e"a(t~T5d7dt; (iii) to take account dependence of risk on elapsed time since exposure. Excess risk a t 2c{t) (t2-t)Itdt The last model is the most general, and, by appropriate choice of k, can take account of all three of the considerations listed above. For example, it reduces to the first model where k 0. For this reason, we apply this model to observed doseresponse relationships. As described earlier, it is a good empirical fit to data on mesotheliomas, but empirical data on the age-dependence of lung cancer probably do not fit it so well. For constant exposure (c (t) independent of t), this model has two important consequences: (i) the cumulative risk after T years of continuous exposure is proportional to Tlc+2. Empirical data on mesotheliomas fit this relationship, with k lying between 2 and 3. 0UC42 42 UCC 000600 (ii) the cumulative risk after T years of continuous exposure is proportional to 1/k+l. Hence, if the first model listed in this section (in which k=0) is used to esti mate cumulative risks, these estimates will be too high by a factor (k+1). Another way of putting this result is that only a fraction 1/k+l of the exposure period makes a significant contribution to the cumula tive risk. For example, for lifetime exposure (T*73 years) and for k*2 or 3, the effective exposure period is only between 18 and 24 years; this result is illustrated in Table 3. For this reason, in this report we relate risks of mesothelioma to cumulative exposures during the first 25-30% of the exposure period. It is less clear that this procedure is appropriate for lung cancer. The cumulative frequency of lung cancers in asbestosexposed workers increases with age in parallel with that in the general population, suggesting that the above model is suit able, with k equal to about 3. On the other hand, the data of Seidman et al. (1979) on workers first exposed in middle age suggest that k should be much smaller than this, perhaps nearly zero (see discussion in Section IV.B and Table 2). How ever, other data, such as those in Table 5, indicate that expo sures to asbestos make little or no contribution to lung cancer risk within the first 10-15 years. Recognizing that a different model may be required to fit the data on lung cancer, we will assume conservatively that the effective exposure period is 00043 43 UCC 000601 equal to about half of the actual exposure period (equivalent to assuming kl for this specific calculation), except that we will exclude from the effective period the last 10-15 years before observation. r i E. Observed dose-response relationships 1. Lung cancer Nicholson (1981) reviewed nine studies of the mortality of asbestos workers and derived 11 estimates of the relationship between excess lung cancer incidence and cumulative exposure to asbestos. These estimates are retabulated in Table 4, with background information on each study and a summary of the basis for Nicholson's estimates of dose-response coefficients in each case. Table 5 includes notes and comments by us on the relia bility of these estimates. The measure of response per unit of exposure used in Table 4 is the percentage increase in relative risk for lung cancer per unit of cumulative exposure. Use of this index is predicated on two assumptions: "(i) for a given level of exposure, after an appropriate latent period the relative risk for lung cancer is independent of the elapsed time between exposure and observation; (ii) the relative risk for lung cancer increases linearly with cumulative dose. Both these assumptions have substantial empirical basis (see Peto 1979 and Nicholson 1981 for discussion). Table 6 tabulates data on deaths from lung cancer among 17,800 insulation workers in the United States and Canada (Selikoff et al. 1979). 00044 44 UCC 000602 4> T rtb e,j f i g u r e t, tei>nr ~ m ejm iie m rn ts t .>l i I t r><) f en |>.?M <K I. - t l h ^: -- <5 c Z *3 2yi ^ <1 < ** 4i,T ' / E *\ I b~ > S ---- 1 * to C W J Stc S' -- - - - to -- L -- t3> >-- w f -- ',1 ; ? t j/ 0 to tX 4 X w to 3w / E X 4 < --U X i'. 3 S ib S " -- * ^ *r m 9< in -- -- tvo 4 c ** t4 v z *-- C n t/i r m -- * <" w e :-- *3 * * --4 0r- --Ij Q., --^ V5 <" S -- to e 9 0 e m mo 4 -- -c W*l A to, e Ml J 3 *-- * ii C >* X - i~ w J<C ttoo 4*mm * r3 j 4 4 3- a.-- X x *-- sto + e ^* o C- 4 J C IfC 0 c U ft*1TM1* *** toi --x--*- a * 1/ 1C3 tIto*oVD a--* TMOi t3to v e vi 0 0 * x to ''to - to to 0 n* -w 4-- a e v e f i 5 S 873 0 o -- t --X w a ti * f 0 -- i* Z ~ X &/M to 0 ** * toi t to If fi ^ if u a -- til in f 4/*h9 C 3- I*f o X'J * If 0 --* % O -- Z 0* e 3 4 *3 Wri lift i* 3 C '- m> *X to V! to - t ; 0 7 w O ee .jj 0 to id jt to* x4 --x if a w to vi 2* 0 t 4 ^ to to > - c to * to 0 -3 - *to -5 ^t C<fl 4 to -- l/l w * s -- Ji T - Q i> E >-- JXM 4 - C -- E^ -- \s*i- ^i O i/i t '.I ^ E S <to 11 w S W111* "to C *- t3o f ^ V. C 4* 4 Wri W J tt 4 0 to to fS tfl to > 4^31/ in .3 to . e. e o t 0 to* M (ft ^ o --in *e9 1/ C to m 0 -s fl -- 4 4 x > _i w -- 4 ii I to to -- ^ Ml * a/ c -i * >* t9ow0 g 0if 0C *to u tIoD uif ttoor.i 4 If to to C X Z 0 2 1 If to 0X to a " y iJ to 0 ^ *9 ^ to ^ 4/1 m3 Of ? to "* to> 0 - *M ^ w- jj J 4) 1 4/ - 0to 4 W e f^ o ** e ^ tori 40 tl) M 9 e s> ^ :* ^ to - VI to ui i> to ii > l/f J* Cu I0 <fl -- w* l -- f0 X 4 4 3 h< 3. 3 "* X* > q X w4 S o to 0 - 4 P- to J 6 3 t0 3",3 3 E to) toi 4 4 toi v m c- e y 5 4 - 3 fl E J til -- o ** 4 lX t X -- 4 X ^ to <9 to M ^ to 4 -- "5 * C to S 4 to. *- > -- It to C 1ft \ ? D y . c to^to w C C5-t> D 0 V) 3T^' XX *f to S U 4 4 *0 to 4 3!I to* to1 4 to O c1 - 0 E U* 0 VO 1. . > to 4 : O4C J 50-; n . u 3 : to K C X 'to to - W to 4 --:4-- o I 3 t 5 to J Ito a 1: w >i 0^ fX > c > wX >* r o 1E U*l K to, t P- ^3 *e * 3 w -3 4 >* " to* w X * >* C -- :3 CX to 4 > to 3 to X *2 rt * E. i-i *f K S V) 4 C to ** * Z to 3 3 6 * 4 0w ^3C l XS 2-= x 'S - X 4f 4 J^ v* !- < < ff* e -! a. 3 --z e to 0 0 3' -- * C 5 to y 0 44X ?34 1 w e to tori e ^ x (A - u 4> *- >-^ to* h* C --X * 3y 4 o i ^ U V -- til 3 to to 3 031 U-- * t O' e --* w= s X to to U * tXo u3 to e3 4^ 3r x wi 1 t3o 1 4-- --U *-- o 4 to> to to) c O 4 -- X to n >* 4 to* 8= Io < y x^ Tef -- i/j to * 35 to 0U-* to 4 * a 9 to* to to 43 4C e> 4 14 -- < e*a 4-- i -- x-- to 3 0 Sw 4 X L -to1 >* 4 -- IH: u a * *3 --eto -xs 3 0 t9o t<oJ ~4 to (J ^ ** 9 S to- -- 4 'P S > 5c s V) w O 1 < fib I X 0 x >, c 0 -4 oto -3 to ^ o cn 4H I- 0.3 (rjir cur 00045 45 UCC 000603 o lih rrv i| * mm fl 2 a fl a 1M fm mm <to m fl m tfol m to w(1 9 01 c * O 0 -- 3 a.-- 1 to fl 3 ^ 2 a. to. m A w fl 01 9 toto01 - 0 fl to 3 m fl 0 fl r mm mm to A J J J= a^ W fl K fl to `to 0 fl a. 0 x - 1- fl fl. -to Oi Uto 2 & t m-- * * C ki I -- a N. -9 * d v -3 > (A 'J 1 A C-* Ci fl -- V; C ^ *J 3 ti r7 d -- cm 3 V0 * -- d--' j: ^f>l fKil 3. A 9 m to 3 3 --5 s* 3 y " *9 --9 Ci >- m e 3d > d3 4to 0 3 * 0 0i d -- -- A > O 3.0 d^ A S Ca 37 9 fl V 01 s 3 - A-- > A tot 9 h 3. 3 * - * f0 3rt d> 3 * V< * f^ 0 c 0 1 fl tofl fl M ^ w C r* 3 y -to 9 j= fl 9 frd to4 to -to d fl -- X ea 9 C fl fl to e a fl 0 1 xs> -- -a fl 7 to >m 0 IN to 0 e o A Jf 9 ii 0tao W S to A --3 * 01 > d * v to 0* (fl -to 3} *3 - * <3 a* d7 --41 s o to * o e a* C 9 <a fl vi * lr U P* X- 9 o S es Q %/ A 7a 9 fl a V to 01 01 d * * to fl > d i >. ^ fl to 3 = 30 to 2. -.1 -- r* v"i * 2 --1 -'Cto3H.CC -3 e 7 4> * fl | 3> d - 7 A J0 C -- 01 A d-- a M te , : ' ' j I a c fl avA n -j 9 to 01d 4-- *9 > to* d>01 3 Vto f vi ca >* \ da to. C 3 9 0 7d - 'iz l * a dd - -3 - . 1 mC fl 9 to 3< > fl V fl 5 fl c a5 a fl d *^ i 3 f3l fl 3 to to d 9 , ' ! 1 ' i : ^ )F : ! to > fl 0 t9o ~e a a >*A 3 fl vi fl fl C fl 2. o*a52 9 V M A f S> -- 0</ c 3. d * fl --* 93 E to -- ^ ua y *r n m id * y 0 3 : S' 3. -9 d *9 C to. fi 3 V dd C C d -- 3 to. 0> > . to 3 a9-- a d SB d 01 at > -- i tf) C 1 5 S '- t ** to | Q3 --V 9 0 3V i r- fl to --to 7-- 01 5 tCot 0) to to "> If B 3 a 01 ; t -- i. e -i V A 3 C-- -- Vto 0fSl d ttoo N-- c9 3 -d ttoo "-- j-i fl to ff 3 ^ Ci---- u fl J fl to to fl to > mm fl to* fl 0 > g a fl to * O -- fl fl 3 fl VI -to i^ fl to d9 Q fl 3 m - *3 fl e 9to fl to ^ fl to to to 3 e * 9 fl * E > a- 3 > -- 1 fl fl B 0 7 fl 1C a fl fl --a9 to > to O fN * to >. w 0 VI 1 0to E M I m a te o f Dim a t Io n o f s tu d y 00046 a3 0 -* n * > I- < 0. S9 0 -to fl to to to U fl > to 3 9 1 fl 9 X 09y ff>l --dX0 0 *0 05 ai fl >to *9 Ae 0 e to fl C UAA ---- 9 3 fl * fl di e C d 4 fl fl fl 9 O to u 9 g3o 7 0C fl tCo --to 9fl e fl c ---- a X u .0 46 o V A 9 9 90 C 7 7e C. 9fl to to fl E* C -- -- fl X u 7 UCC 000604 TABLE 5 NOTES ON STUDIES SUMMARIZED IN TABLE IV-4 AND ON NICHOLSON'S DOSE-RESPONSE ESTIMATES Study No. and Reference Notes and Comments 1. Seidman et al. 1979 This is the most useful set of data because the workers were exposed to asbestos for only a short period, so that the dose data are not confused by late exposures. However, the dose- response relationship is not a very close fit to a linear relationship (Nicholson's Figure 4). The point for the highest exposure group (ca. 140 f-yr/ml) falls below the linear regression line, probably due to dose saturation (Nicholson, p. 15). The points for the three lowest exposure groups (less than 7 f-yr/ml) fall above the regression line. If the calculations were based only on these three points, the estimate of % increase in lung . cancer risk per unit dose would be 37% per f-yr/ml. However, it is possible that the asbestos workers smoked more than the general population of New Jersey with which they were compared. This would explain the elevated rates of lung cancer in workers exposed for only a short period. If the regression line is not constrained to pass through a relative risk of 1 at zero exposure, the estimate of % increase in lung cancer risk per unit dose would be about 2.4% per f-yr/ml. 00047 47 UCC 000605 TABLE 5, Continued Study No, and Reference Notes and Comments 2. Newhouse and Berry 1979 , These estimates are "necessarily uncertain because of the limited information on both exposure and length of employment" {Nicholson 1981). Nicholson assumed an average of 10 years for those em ployed more than 2 years. This may be too high. The estimates of cumulative exposure for males are probably too high because some were followed for only 11 years after onset of exposure. 3. Henderson and Enterline 1979 This study was limited to men who had retired at age 65 and hence is likely seriously to underestimate the effects of asbestos exposure. Although Nicholson {1981, p. 28) incorporated a small correction for mortality prior to age 65, he also showed that Henderson and Enterline's study failed to detect a large incidence of mesotheliomas at the same plant. The effective average dose would also have been overestimated because of the inclusion of workers who had started work as little as 6 years prior to the end of follow-up. 4. Nicholson et al. 1981 The estimate of fiber concentration for the low-exposure group (10 f/ml) is tenuous, although the average expo sure estimate for the entire factory ' is close to that of Henderson and Enter line (1979). Even if the concentration is correct, the estimate of cumulative dose (apparently 270 f-yr/ml, or 10 f/rol for 27 years) is probably much too high, both because such a dose would include exposures within 10 years of the end of the study, and because it would be expected to lead to dose saturation effects. 00048 48 UCC 000606 TABLE 5, Continued Study No. and Reference Notes and Comments 5. Dement et al. 1981 ' 6. Peto 1980 This study appears to be based on good exposure data, involving calibrated measurements throughout the exposure period. However, the number of lung cancers recorded was small (26 versus 6.5 expected). The exposure estimates included some within 10 years of the end of the study and hence are likely to have overestimated effective dose levels. This study is difficult to interpret because of the major revisions in the exposure data, the small si2e of the cohorts, and the unexplained difference in mortality patterns between the workers employed before and after 1951. The early cohort is anomalous because of the low rate of lung cancer (relative risk 1.5) despite a high frequency of asbestosis. Most of the workers in the later group were followed for less than 20 years after first employ ment, so the effective dose is likely to have been greatly overestimated. When the cohorts were considered by period since first exposure, the rela tive risks ranged from 0.45-5.34 with the 0.45 risk in the subgroup employed before 1951 with 35+ years since first exposure; and the 5.34 risk in the subgroup employed after 1950 with 20 years since first exposure. COO 49 UCC 000607 TABLE 5, Continued Study No. and Reference Notes and Comments 7. Selikoff et al. 1979 This estimate appears low for at least three reasons. (1) Actual measurements of asbestos concentration were in the range 3-9 f/ml; Nicholson's estimate of 15 f/ml is based on assumptions of greater exposure in the past. (2) Nicholson assumed exposure for 15 years, which is probably too long for contri bution to effective dose. (3) At such doses, dose saturation effects would be expected. 8. McDonald and Liddell This estimate may be low because some 1979 workers may have been employed for as little as 10 years at the end of the study. Nevertheless, a good linear dose response relationship was obtained (Nicholson 1981, Table 8). This study also yielded anomalously low figures for incidence of mesothelioma and asbes- tosis, considering the very high dust levels. ` 9. Nicholson et al. 1979 The estimate of fiber concentration is tenuous, and the assumed period of 20 years exposure is probably much too long to contribute to effective dose. Although the estimate of risk per unit dose is 2-3 times higher than that of McDonald and Liddell (1979) it is still an order of magnitude lower than that in most other studies. This may reflect lower risks from mining asbestos than for the other processes that have been studied. 00050 50 UCC 000608 Although the number of lung cancers per 1,000 person-years in creased rapidly with elapsed time since start of employment, the relative risk did not vary markedly, at least after the 25th year. Table 7, based on data from Seidman et al. (1979), similarly shows that relative risks were more or less constant in each exposure category after about the 20th year from onset of exposure. Figures 1-5 illustrate the dose-response relation ships compiled by Nicholson (1981) from data available from five independent studies. In at least four cases, the data can be fitted closely by a linear dose-response relationship. In the fifth case (the study of Seidman et al. 1979, illustrated in Figure 1), the curve rises rapidly at low doses and flattens off at the highest exposure level (interpreted by Seidman et al. as a dose "saturation" effect); the three points for the lowest exposures fall significantly above a fitted straight line. This may indicate that this curve is not linear, and that incremental risks per unit of exposure were higher at low doses (see discussion in Table 5, note 1). With these limited exceptions, epidemiologic data support the assumptions that relative risk is independence of elapsed time and is linearly dependent on cumulative exposure. Hence the slope of the linear relationship (incremental increase in relative risk per unit of cumulated dose) is a stable measure of the incremental risks of asbestos exposure, and may be used to characterize the results of each study. 00051 51 UCC 000609 2 Ps a > a c u < < f9> <c<z< --< *K u 23 < j: * to ma ttn e 2 fr* * H xC < mx c-- *- ^X U x <> o3 4 aX ito r* to O H c x v* 3X 5 (SI zw 1- v <to 5 5 (/j i i 1 wI 'J i 41 'J Cy * -- -- r* i/i a u OOOOOOOOO & ? ai tio c* a e o to tfl i *to ut a i* ^ 0* * -- 1 e -Ni/'MV'*'- | C OOQ..3.--..< ! I ---- ia S -c i ,/ *t o r* ^ o ^ 4) T Vft * W 1 * f C j( -- V) a&t --* 0 44J) u">s*oaa<Ba`e* VT9Q490 *ft OfNrn^/^lAlnT',,' i 41 (toII i* c t*o 9> Se ei Miso^e^e -3 \ ji u 4 c 3 3 ttoo i> a. it 41 vc >a e zo 1 | I | > J e 4> 9 > 10 9 tl1 i *e *-0^^snO- g V> i e ti l zt i - to * 41 | .*to z</| Z 0- 1 i e 0 * i ' to to tmo 9 0to V VI c ^enet*fl *r ! Q^ 0 coc 4> ** 52 UCC 000610 TABLE 7 RELATIVE RISKS FOR LUNG CANCER BASED ON CUMULATIVE PROBABILITIES OF DEATH AT DIFFERENT ELAPSED TIMES FROM ONSET OF WORK IN AN AMOSITE ASBESTOS FACTORY. (RETABULATED FROM DATA IN TABLE 6B IN SEIDMAN ET AL, 1979) Length of time worked <1 month 1 month 2 months 3-5 months 6-11 months 1-2 years 2+ years Elapsed time (years) since onset of work 5-20 5-25 5-30 5 -35 0 2.05 1.42 1.27 1.78 1.67 3.11 0.74 1.60 1.72 1.76 1.84 2.14 2.63 0.77 1.43 1.82 1.69 2.00 2.89 2.78 1.24 1.47 1.58 1.76 2.01 3.20 2.79 00053 53 UCC 000611 FIGURE Es limn led Dose of Asbeslos ( fiber - year / m l) u aM E-1 r~ E S' EM" wc < c Zc < SX <i--i WE* i--2 oE- Oi- c 3u VfCaJ >-- *4 ^ * -4 fl fa) CJ c *f6a u CJ z < V u2 J C2 2** CCJ u< X w < >> ap.f:i OJ 4E0f4a f0al Vi 2 fl Ocs fa. X a Xv p+ El. Z o I* 3 Os <efaC Z<w*. Vi rO* *4 c fa.a b. VMi O 0 * y ac a Vi < *04 0 VMi X t* cs- z s w C2 W tr > => w E* Vi * E* < O < M 2 c. X faj fa.W tt W 2 w 3 C M H o o OJ o o o CD o CO o o OJ p*. CO in ^ fO oj -- o pspsdx^ /paAjasqo J3DU03 6un"| UJOJJ m03Q ;0 >lSiy SAi;ojsVf i 00054 UCC 000612 0055 RELATIVE R ISK OF DEATH FROM LUNG CANCER AMONG WORKERS IN A U .S . T E X T ILE M IL L , ACCORDING TO THE ESTIMATED F EXPOSURE ig u re 6 in TO N ASRESTOS ic h o ls o n IN FIHER-YEARS/Ml (re d 1901; d a ta from Dement raw et na l .fr om 198 ps;:sc!x3/p3Ajssq0 jsoudq bun-] UJOJJ U'DSG p >iSl 0 A11D i e ^ 55 UCC 000613 50 100 150 bslim oted* Duse of Asbesto ( fiber - yours / m l ) C0C5S oc 0} 0 JZ u-- 2 r- C -* (3 SuJ -j sa H <3 t*-l 'J 3<u5 =0u ua 5 GO c> jsouoq Ajc;DJidssvj JOJ. 3AljD|3^ 56 UCC 000614 R E L A T IV E R IS K OF DEATH FROM LUNG CANCER AMONG WORKERS IN A QUEBEC ASBESTOS M INE AND M IL L ACCORDING TO CUMULATIVE 1000 2000 3000 Cumulative Dust Exposure (mppcf years) o CM o RE1ATTVE RISK OF DEATH FROM LUNG CANCER AMONG RETIREES FROM A U .S . ASHBSTOS PRODUCTS F A C IL IT Y ACCORDING TO CUMULATIVE 200 400 600 800 1000 Cumulative Dust Exposure (rnppcf years) OD C rs 5 --. 0 r- .c <r. Hu -i Z &C C Li r- --o OC Li X, C'-C - c i. (0 Eou ccm U-i Li It - XJ *u3 --c0 CUl E0 -- Li tt D'K-0 -<fU5l C XJ X w 6- Ui z> a -- co cD c c\j J30UD3 XjoiDJidsajj J0j 3A!lC|0^ O o 0.0057 57 UCC 000615 FIGURE lis lirn o le d Done of Asbesfon Ci < E<" w s aa-- <zH 0L)l t--3CfaT--tcC- Xasc 2O Oz o z< sc u u<z tEyo" cE" CzJ asc <o*yw1 VE0LU c 5iLs --0uD1 -4 z LNLii S0) "C3 <0 30)) 30 J= a czz Z o b. M >sc Co* uC<b to 3uEa" sCO SC z>0 < E >1 0Li 0C u-i tiC aM Ua Eb m "O UE< o CL z a b. O lt-ol to tuECo> tCOEoO" aCco < -c4 sc0i -4 sc < 0 u > 44 e*<*4 U1 ft-r4* o E* suc => CO .= w04 z c ~4 u sc s oc. X U o o m o o o o J30UOQ 6 U n-J ULIOJJ q|D9Q p )\s\# 9A|p|3y ; 00058 58 UCC 000616 Table 4 shows that estimates of this quantity vary over a 150-fold range, from 0.06 to 9.1% increase in lung cancer risk per fiber-year/ml of exposure. However, 6 of the 11 esti mates fall in the more limited range between 1.1 and 9.1% per f-yr/ml. As shown in the notes in Table 5, each of the five i lower estimates is anomalous in one way or another: (i) Study No. 3 {Henderson and Enterline 1979) was based only a study of retirees older than 65, and there is independent evidence that it missed a substantial fraction of asbestos-related deaths. (ii) In Study No. 6 (Peto 1980, and other reports), there was an unexplained difference between early and late cohorts of workers in the same factory. The early cohort had an anomalously low frequency of other as bestos-related diseases, and the later cohort was . . not followed for long enough {maximum 25 years after onset of exposure) for adequate measurement of effects. {iii) Although studies 8 and 9 (McDonald and Liddell 1979, Nicholson et al. 1979) both involved uncertainty in estimating dose, this is unlikely to explain the ex tremely low estimates of risk per unit dose. Both studies also reported anomalously low frequencies of other asbestos-related deaths. Both studies were of workers exposed during mining and milling operations, raising the possibility that risks are in fact much lower in these operations than in those involving manufacturing and application of asbestos products. A review of the response of these workers to cigarette smoking also showed the possibility of substantial underreporting of lung cancer mortality, or very large competing risks that killed the potential lung cancer cases at earlier ages. For these reasons, we place primary weight on the dose- response data in the range between 1 and 10 percent increase in lung cancer risk per unit of exposure in fiber-year/ml; al though several sets of data yield substantially lower estimates, these estimates are either questionable or of doubtful relevance to general population exposure. Within the range of 1-10% per 00059 59 UCC 000617 f-yr/ml, there appears to be substantial variability from study to study. Some estimates may be too low because of overestima tion of effective doses, and the highest estimate is uncertain because of deviations from the linear model (see notes to Table 5). Data presented by Hammond et al. (1979) show that asbestos and cigarette smoking act synergistically in increasing the risk of lung cancer.^ Compared to nonsmoking controls, smokers not exposed to asbestos had a relative risk for lung cancer of 10.85; non-smoking asbestos workers had a relative risk of 5.17; and smoking asbestos workers had a relative risk of 50-55. These data can be used to estimate lifetime risks separately for smokers and non-smokers. In the United States population as a whole, the lifetime risks for men of developing lung cancer are about 0.1 for smokers and 0.01 for nonsmokers; for U.S. females, the corresponding lifetime risks are about 0.04 and 0.005, respec tively.2 Hence, the estimated increase of 1-10% in these rates ^It has been suggested that the major cigarette-smoking effect . may be to move forward in time the age at which the lung cancers occur. This suggestion says that followed enough, the nonsmokers would eventually develop as much lung cancer as the smokers. 2 If the lifetime risk for nonsmokers is r and that for smokers is Sr, and if the proportion of smokers in the population is p, then the average lifetime risk for the population R is R * pr + (l-p)Sr. For U.S. males, R*0,0595, S*10.85, and p is assumed to be 0.5 for the period 1940-1960 (data from Hammond et al. 1979 and Schneiderman et al. 1979). Hence, r is approximately 0.01 and Sr is approximately 0.11. For U.S. females, R*0.01559, S3, and p is assumed to be 0.3. Hence, r is approximately 0.005 and Sr is approximately 0.04. 60 O'JCi'O UCC 000618 per f-yr/ml exposure to asbestos would lead to the following increases in lifetime risk of lung cancer^": male smokers: 0.001 to 0.01 male nonsmokers: 0.0001 to 0.001 female smokers: / 0.0004 to 0.004 female nonsmokers; 0.00005 to 0.0005 According to the model used here, these risks would be more or less independent of the age at which the exposure took place, provided that it started at least 10 years before the age (about 55 years) when lung cancer incidence begins to become large (see Peto 1979) . 2. Mesotheliomas There has been little published discussion of dose-response relationships for mesotheliomas, and the following analysis has been developed for this report. Pleural and peritoneal mesotheliomas are considered together in this analysis, since ^These estimates do not depend on the prevalence of cigarette smoking among asbestos workers, because all the calculations of relative risk underlying the data in Table 4 involved com parisons with lung cancer rates in the general United States population. According to Hammond et al. (1979) about 83% of asbestos insulation workers in the United States and Canada smoke (or formerly smoked) cigarettes. Even in the absence of asbestos exposure, this would increase their relative risk for lung cancer to about 1.6. This may explain the anomalously high relative risk for highly exposed workers in the study by Seidman et al. (1979), which may have led to an erroneously high estimate of the dose-response curve (Note 1 to Table 5). However, this is apparently not a general problem, because relative risks as low as 1.1 and 1.27 have been reported in workers with low exposure in other studies. 61 UCC 000619 their patterns of occurrence are often (although not always) parallel. Since mesotheliomas are rare in the general population unexposed to asbestos, figures for relative risk cannot be cal I culated. In groups of workers exposed to asbestos, the incidence rate of mesotheliomas increases rapidly with age and duration of exposure, at least until the age where other causes of death become important (Table 6, Newhouse and Berry 1976). Where comparisons are possible, the incidence rate of mesotheliomas increases more rapidly with time than that of lung cancers (Table 6). Hence, the ratio of mesotheliomas to lung cancers increases with time, approaching one in some exposed groups towards the end of life. Because the frequency of mesotheliomas varies so strongly with time, it is often difficult to obtain comparable measures of response from different studies. In most cases reviewed below, data are available on the number of mesothelioma deaths in the study groups 35-40 years after the onset of exposure. However, these measures are not always strictly comparable: they may significantly underestimate effects in groups within which exposure was started at different times. Tables 8 and 9 summarize data from two studies in which the frequency of mesotheliomas was recorded in several groups of workers with different degrees of exposure. Despite the limited number of mesotheliomas and the uncertainty in exposure COO 62 62 UCC 000620 TABLE 0 DOSE-RESPONSE RELATIONSHIP FOR PLEURAL AND PERITONEAL MESOTHELIOMAS IN AMOSITE INSULATION WORKERS, FOLLOWED FOR 35 YEARS AFTER A LIMITED PERIOD OF EXPOSURE. Length of time worked <3 mo 3-5 mo 6-11 mo 1-2 yr 2+ yr Cumulative dose (f-yr/ml) 5 No. of workers 223 No. of deaths from mesothelioma between 5 and 35 yrs after onset 0 % of workers with mesothelioma 0 % occurrence per f-yr/ml 0 12 149 0 0 0 26 52 140 125 125 188 3 2.4 0.09 4 3.2 0.06 7 3.7 0.03 (Compiled from data of Seidman et al. 1979, using estimates of fiber concentration from Nicholson 1981.) C00S3 63 UCC 000621 TABLE 9 DOSE-RESPONSE RELATIONSHIP FOR PLEURAL AND PERITONEAL MESOTHELIOMAS IN AS3EST0S PRODUCTS MANUFACTURING WORKERS, FOLLOWED FOR UP TO 42 YEARS AFTER VARIABLE PERIODS OF EXPOSURE. Concentration of dust in workplaces: Low/moderate High Period of work; Nicholson's estimate of cumulative exposure (f-yr/ml): <2 yr >2 yr 7.5 75 <2 yr >2 yr 30 300 :ales (11-42 yr followup from onset of work) Concentration of dust in workplaces: Low/moderate High No. of men No. of deaths from mesothelioma % dying with mesothelioma % occurrence per f-yr/ml 884 554 . 47 0.45 1.26 0.06 0.017 937 16 1.7 0.06 .512 . 19 3.7 0.013 Females (33-40 yr followup from onset of work) Concentration of dust in workplaces: Low/moderate High o. of women No. of deaths from mesothelioma % dying with mesothelioma t occurrence per f-yr/ml 98 1 1.0 ca. 0.03 396 13 3.3 0.11 199 7 3.5 0.012 (Compiled from data of Newhouse and Berry 1979, using estimates of cumulative exposure from Nicholson 1981.) 00064 UCC 000622 levels (see note 2 to Table 5), all the estimates of response per unit dose were of the same order of magnitude, between 0.012 and 0.11% per f-yr/ml. Although this is not strong statistical evidence for the hypothesis of linearity of dose-response rela tionships, the data are consistent with this hypothesis. In the remainder of the analysis, the assumption of linearity will be made, and the measure of response per unit dose that is pre sented will be the percentage of occurrence of mesotheliomas (through about 35 years after onset of work) per unit of exposure (f-yr/ml), with the recognition that duration of followup is of considerable importance. Table 10 summarizes data from seven studies which permit such estimates. As in the case of lung cancer, the two studies of mining and milling operations yielded much smaller estimates than all the others. In the case of mesotheliomas, the discrep ancy is so large (by two orders of magnitude) that it is diffi cult to explain it away as due to methodological problems: it suggests that asbestos from these operations may be much less hazardous than that from manufacturing operations and appli cation. Despite the considerable uncertainties in the calcula tions, all the other estimates fall within a reasonably limited range, 0.005-0.06% occurrence per f-yr/ral. Because the lowest of the figures is based on a very small sample (and derives from a study which yielded an anomalously low risk for other * cancers also) , we suggest that the range most relevant for pre dicting risks in other situations is 0.01-0.06% occurrence per 00065 . 65 UCC 000623 4 w 4 e 4 to 'a. K O b W5 35 < H id ^ a 2 9* H-- Ic x to * to c 0x 1- * > X to 4 to - i4&5 "1 -- 41 X to ^ x0 y H & *.T- i to * X4 to i S Z4 O HT tko 4 4 9 Cy ze <- X4 < to to O XV to to 0y Kf Vtoi v a to Z4 M 4 VJ r u 9 43 to -- QC 4 to 3 z a a tCo t4o <x Z4 r 1S0 to * ; : 4T ' 1 4S 1 4 94 : x 4 to 1 - 4 4 1 400 e 4 to 4 Z to a 0 1 <4 \ j* * y. \ N to to tao IN I 1y 1& i 4 to -- 1 --a-- j e c-i f 5 D N i & 4m 1 4 to 4 >1 ! iU: i >4 1 ^> -- * 4 -- 4 to 1 to to w 1 4 4 5 N. 1 1 -- V. to ; r 1 =.* n!i. to 'J to 'to yt ee 0Q tomt p% j *y i 4^ ~ > * ; mC4 ! 4 i X Ji *M ^ ! = rc f ; 1 i lz to -- to a to a *1 N V (N M to to mm w i/S O 4 N t"o -- C3 M 9^ ?9 4M 3 4 2 S >1 K to J " f--* mS 4 0? 0 to 3 3 to to w 34 VI 9 4 3 <b< 3 coo to xto cc 0c K1 a* VA M s to 4 K9 M 3 *4 > 3 X <N 4 4- 0 0- 4 44 IO V4 1- < a *5 s 3 m 0Z to (9 4 -- 5 0 t < 7* e c to Q3 to e to u 4 4 4j mm to ^ 3 3-- 4e4 e 4 VI m t-- 4 \\ M\ \SS ^ id *'v ii<to e,/> o -r* *a \ Q X l? J S X IJs s. j*: ? ; iX*X * 4-- -e o* 0 >* 0 to 0to to U3 ^ QC -- to 01 4 cy t4o >to ^ 4x e CJ -- Q 4 4 44 -- to to t00>o4 5y f X U0 t0o to0 : to y 40 K4 943 ttt00o4oo > to X u * cfl to 1/S'- > --et c4j <** t4o tt00oo t4o to >n 0to g 0X ! u- t4o t00o > to X0) * o 0 4 > 9e r* os 9 4 > -- f w i U 3 y J 4^ * 1 c 5^ to 4 X s & -- 4 to e -- 9 C I to Vf 9s J i 5*^ il to 4 X to * --4 SS ! a 4 4 to to 1W * 1 *39 *3 0 , to J3.-^ oavto. to c c S 41 0 ^ to- ^ to to to- 9< j 3 to tyo to toos E9 atot4tto -* to ^ to - -* e* r* * ^ to 9< 4 to c------C9c"- stteooS5ye 2"g -- i-j 0* . --c S0 'r p* n 1 to to 1 Tc----rC s-c3 --i*u.. Z 4 to. o> c & I s 1 tVo 4 4 4 4S X 71 if X 0 V ft 9 4 'J 4 X V 4 4to1 4 a 5 9* w99 4 - v4 to t440o 4 to 44 XT'X4 to 4 X4 40 4 y0 V3 a4 00066 66 UCC 000624 f-yr/ml. This represents the cumulative risk to about 35 years after onset of exposure. The data in Table 6 suggest that risks continue to rise rapidly after 35 years; for the workers whose experience was summarized in Table 6, more than 80% of the life time risk occurred after the 35th year. The last column of Table 10 gives the ratio between the number of mesotheliomas and the number of lung cancers in each study. This ratio has varied widely, from 4% to 100% in differ ent groups of workers. The data in Table 10 suggest that this ratio may increase with increasing exposure, suggesting that the dose-response relationship for one or the other type of cancer may deviate from linearity. It is not clear whether the dose-response function for mesotheliomas tends to curve upwards, or that for lung cancers tends to curve downwards, but the data in Figure 1 suggest that the latter may occur. 3. Conclusions . (i) Data for both lung cancers and mesotheliomas are gen erally consistent with linear dose-response relation ships. Several sets of data for lung cancers agree closely with linear relationships. {ii} After exposure to asbestos, incidence rates for lung cancer increase with time in a manner generally par allel to that of rates in unexposed controls, so that relative risks are more or less constant. (iii) Incidence rates for mesothelioma increase with time more rapidly than those for lung cancer, and may equal them 45 or more years after the start of exposure. (iv) For lung cancer, excess risks per unit of exposure vary widely, but estimates are clustered between 1 and 10 percent increase in cancer risk per f-yr/ml. Lower figures derived from some studies may be explained by difficulties in estimating exposure or by variations in exposure situations. Some variability in apparent C0067 67 UCC 000625 dose-response relationships may be due to the confound ing effect of cigarette smoking. (v) For mesotheliomas, most estimates range between 0.01 and 0.06 percent (cumulative risk after 35 years' exposure) per f-yr/ml. Two estimates for mining and milling operations are lower than this by two orders of magnitude. Depending on the age at first exposure, lifetime risks may be five or more times higher than this. ' (vi) Except fob the apparently much lower risks resulting from exposure to asbestos (chrysotile) from mining and milling operations, there is no clear evidence that risks depend strongly on the type or formulation of asbestos. (vii) Lifetime risks for workers exposed to 1 f-yr/ml are expected in most cases to lie within the following ranges: Lung cancer (male nonsmokers): 1 - 10 x 10 -4 Lung cancer (male smokers) : 10 - 100 x 10 -4 . - - _4 Lung cancer (female nonsmokers): 0.4 - 4 x 10 Lung cancer (female smokers) : 3 - 30 x 10 -4 Mesotheliomas (first exposure at age 35): 1 - 6 x 10 -4 Mesotheliomas (first exposure all age 20): 4 - 20 x 10 (viii) Although the above are the most appropriate ranges to use as the basis for estimation of environmental risks, it is recognized that several studies suggest that risks may be substantially lower in appropriate circumstances. 68 UCC 000626 V. EXTRAPOLATION TO LOW EXPOSURE LEVELS The problems involved in estimating risks to the general population exposed to low ambient levels of asbestos were dis cussed in the introduction to Section IV. Direct measurement of low-level effects in the general population is generally impracticable, and has not been attempted, so that it is neces sary to estimate low-level risks from data on workers exposed to much higher levels. Such extrapolation involves at least three separate steps: (i) extrapolation from high to low exposure levels; (ii) extrapolation from exposures during a working lifetime to exposures throughout life; . (iii) extrapolation from the active, primarily male, working population to the general population. A. Empirical and theoretical bases for low-dose extrapolation models The principal question at issue is whether a linear, non threshold dose-reponse relationship can be assumed to be valid at low exposure levels. Several empirical and theoretical considerations bear on this question. (i) As discussed in Section IV, at least five sets of empirical data indicate that dose-response relationships for lung cancer and mesothelicma in exposed workers are consistent with linear dose-response relationships over a wide range of cumulative doses, from 900 f-yr/ml down at least to 1.5 f-yr/ml. cock? 59 UCC 000627 (ii) Tarter (1981) has recently analyzed data on gastroin testinal and other cancers in the San Francisco Bay area of California ,in relation to asbestos fiber concentration in drinking water. Using sophisticated statistical procedures in the field of pattern recognition, Tarter found a relationship between cancer incidence and exposure that appeared to be linear down into the lowest range of exposures (less than 30 fibers/ml). Although the procedures used were not fully specified in the published paper, Tarter reported extensive testing of alternative treatments of the data and found that all treatments yielded linear nonthreshold relationships. Although this study deals with cancer at different sites, it demonstrates that asbestos can induce cancer at low (environmental) levels of expousre. (iii) It is now widely accepted that the development of cancer is a multistage process, and multistage models are gener ally preferred in low-dose extrapolation, unless there is compell ing reason to use another model in any specific case (Armitage and Doll 1961, Crump et al. 1976, NAS 1977, 1980). Extensive discussion of the multistage and other models has been published by the Safe Drinking Water Committee of the U.S. National Academy of Sciences in two recent volumes of Drinking Water and Health (NAS 1977, 1980). The multistage model is preferred for five main reasons: (i) its generality; (ii) its consistency with a wide variety of plausible biological mechanisms; (iii) exten sive empirical evidence for multistage mechanisms in carcinogen esis; {iv) the generally good fit between the multistage model corto ' 70 UCC 000628 and empirical data; and (v) its conservatism. At low doses, the multistage model generally predicts a linear relationship between cancer risk and dose. Such a relationship is more generally required by any biological model in which the effect of an agent is to add to a biological effect that is already taking place (Crump et al. 1978). It is important to note that the multistage model predicts linear, nonthreshold doseresponse relationships for agents that act at any stage in the development of cancer. (iv) It is now widely accepted that in many or all cases, the first stage ("initiation") in the development of cancer involves an irreversible change in the genetic information in a cell. This change often involves change in a single' base in DNA, so in theory it can be effected by one or a very small number of molecules. Thus, at least for "genotoxic" carcinogens ("initiators"), there are theoretical reasons for expecting a nonthreshold relationship. As yet little is known precisely about the mechanism of action of asbestos. Although there is no evidence that asbestos interacts directly with DNA, it may carry genotoxins or facili tate their entry into cells. At least for mesothelioma, there is evidence from the form of response-time relationships that asbestos acts an early stage (Day and Brown 1980). (v) It is often stated that thresholds are to be expected for agents that act at later stages (including "promoters"), However, there is little or no empirical evidence for this 00071 71 UCC 000629 assumption, and the theoretical arguments for it are limited. The most plausible mechanisms by which threshold-type dose- response relationships could arise are the following: (a) cellular thresholds for physiological changes (Diranan 1972); (b) metabolic overloading (Gehring and Blau-1977); (c) overload f of repair mechanisms (Cornfield 1977, 1980). Apart from theoret ical and practical objections to nonthreshold models based on these mechanisms (NAS 1977, 19C0), none of them could be applied to asbestos. Unlike chemical carcinogens, for which reduction of the dose administered to the animal leads to propor tional reduction in cellular doses, asbestos is a discrete material in which the fiber is large compared with the cell. Hence, reduction in intake by the organism is expected to lead to reduction in the number of cells in contact with asbestos, but not to a reduction in the dose to any target cell. For this reason, a linear nonthreshold dose-response relationship is likely for asbestos, regardless of the mechanisms of its action at the cellular level. - For these reasons, we assume generally that risks to the general population at low exposure levels can be estimated by simple linear extrapolation from the dose-reponse data observed in workers. B. Extrapolation to Lifetime Exposures The dependence of r*sks on age and duration of exposure has already been discussed in Section IV-B. For the reasons discussed therein, we assume that: C0072 72 UCC 000630 (i) for mesotheliomas, the risk per unit dose to a person exposed for a full lifetime (0-73 years) is approximately eight times that observed in workers exposed for half a lifetime (e.g., 20-56 years) (see Table 3 for illustration); (ii) for lung cancers, the increased risk per unit dose to a person exposed for a full lifetime is approximately equal to that observed in workers exposed for shorter periods (see Table 2 for illustration). For the reasons stated in Section IV.D, we assume that the effective exposure period for the general population is the first 20 years of life for the induction of mesotheliomas, but the first 36 years of life for the induction of lung cancer. C, Extrapolation to the General Population There is no empirical basis for the extrapolation of dose- response data from worker populations to the more heterogeneous general population. The limited data on occupationally exposed women show no clear differences in sensitivity from men with comparable exposure (Tables 5 and 10). The existence of a "healthy worker effect" in some studies (e.g., Selikoff et al. 1979) suggests that the general population may be somewhat more suscep tible to lung cancer, inter alia, than asbestos workers, but the difference is small compared to other uncertainties. There is no information on the susceptibility of children, except that they are known to respire a greater volume of air per unit body weight than adults. In the absence of definitive reason to the contrary, we will assume that the general population 73 UCC 000631 is as sensitive as the workers. We will assume that cigarette smoke acts at a late stage in the development of lung cancer (Day and Brown, 1980), so that data on smoking asbestos workers can be applied to persons who are exposed to asbestos from birth and who start smoking in adolescence. i D. Estimates of "Virtually Safe" Exposure Levels The data and assumptions discussed above can now be used to derive estimates of risks likely to be experienced by indivi duals in the general population who are hypothetically exposed to a constant airborne concentration of asbestos throughout life, frcm birth onwards. These estimates of risks are tabulated in Table 11, in which separate estimates are presented for lung cancers and for mesotheliomas. The estimates for lung cancer are further subdivided according to sex and smoking habits, to utilize the separate estimates dervived in Section IV.E. The third column in Table 11 tabulates estimates of lifetime risks for individuals exposed to a cumulative dose of 1 f-yr/ml of asbestos, starting at birth. For lung cancer, these are . the same as the estimates for exposed workers derived in Section IV. E. For mesotheliomas, we assume that lifetime risks would be about 8 times higher than the risks observed in workers followed for half a lifetime after first exposure. Since the estimates are derived from studies of occupationally-exposed workers, they are appropriate for exposure to a concentration of 1 f/ral for one working year (i.e., about 40 hr/wk x 50 wk/yr) . C0G74 74 UCC 000632 The last three columns in Table 11 tabulate estimates of "virtually safe" exposure levels, i.e., the average concentra tion levels which would be expected to correspond to an excess lifetime risk of 10"fi (1 in a million). These estimates are derived by dividing the cumulative dose of 1 f-yr/ml by the effective exposure period (36 yr for lung cancers, 20 yr for t mesotheliomas, as discussed above), and adjusting for expected differences in the duration of exposure each week. The fourth column presents estimates appropriate for outdoor, nonoccupational exposure; the fifth column presents estimates appropriate for indoor exposure, including that in the heme; the sixth column represents estimates appropriate for both indoor and outdoor exposure. Data on the average amount of time spent indoors and outdoors by persons in the FRG have been tabulated in a review by S2alai (1973). However, this review did not include specific data for children and adolescents, the age-groups whose exposure contributes most to lifetime risk. In the absence of precise data, we have assumed a figure of 20 hours per week for the average period spent outdoors by persons under the age of 20. All estimates of "virtually safe" exposure concentrations in Table 11 are given in units of fibers/ml. For comparison with reported concentrations of asbestos in ambient air, which are usually reported in units of ng/ml ambient air, it would be desirable to express them in these units. Appendix A sum marizes a number of estimates of the conversion factors between 75 iO UCC 000633 R IS K ASSESSMENTS FOR THE GENERAL POPULATION EXPOSED THROUGHOUT L IF E TO CONSTANT LOW CO NCENTRATIO NS OF ASBESTOS AW OU pH 0 <w h 0V tu u 3 --' 0) pH e 0a Sx thl 9} >1 C -H 0 -X 4 9) 4J 9) <0 S u Jj C C0 <D 0 *3 Ca 0a oq 43 u 9IN jD 49 -tUp -0UJ hi 3 0) 9) jj E 93 E -4J 9) jj jj 01 -- M --t tp JC CO VO l"J w A 03 iH n u JC (N 93 93 E hi 3 JJ jj W x 9) 0 jj Hvu a -43 xp 93 E jj \q 13 E u a> jj q 5 tj UJ c> Oi 1 <U -u E 49 --< jwJ jj 0) U **0p1 u O JJ a JSJ3 *3 c <0 --ae4 Jaj 49 --t X 0A W93 E to fi y a9>3ti*J 9OCA3 e- 0 0 tO1 to1 O pH opH i1O p4 11 pOH X XX X P* o o o 1 r* <N e r** i 1 <w c* PM i o m to11 o "4 X CD t m JC to to o i o pH i o pH 1O pH t o l-t XXXX o e vo o t tN o 00 i | *N * 90 in 1 o o rsi rti to 1 poH X Ot 1 iH JO to to tO 1 opH 1 Q H io pH t o *4 X XX X oeoe to o o 1 to VO 1 o (N O1 mpH 1 vo <N o mpH to 1 poH X to l o pi < 1 111 a oOO pH i--i *4 *4 XXXX oopH 1 opH o *4 p-14 o ml n 1 o t pOH X Q in i 00 a ij aw 4993 o a u 93 49 0 E ac aw a 49 0 aE .ax 0 aE c 0e En 0e paH paj ppHH < p9H3 q 93 4K-3 q aE q aE X X h. Sy in Esaq t -J c3O' O' c3 O' c3 O' c3 O pH aa aa JJ X *3 76 o a * 0 -pHH 0 a tM u 3 a ta h oa 4a9 X hi a0 3 jj aW -CU H vu *3 a jj > qu a aa Ot-Cl q0 vu jj 0q JJ JJ C aa 3 a CE 91 --j Jj O JJ X aa *3 ee jj c au 91 JQ a aa jQ 0 jj 0a jj 3 u a3 ab a 3g 3 a a a0 X aqa a q a jj u q g ahi a 0 C q o ~PHt a --th4i hi ua a pOcsH' jj 0aa E IjUj q a U hi EU V0U 4a9 VoU 49 a JJ U1u9 u u IQ 01 e OJ taUJ IQ 4J J0J9 x: tu tu u J -H -4 M0J u 0 tu a u q a > to n w g 0 w 0) x; jj 0 w 4 S u 0 tu a u 3h Q <N a 0 JJ 'Da 3g aa q 'S 0u U aa Jj X a a jj u3 a a C2u ''aa* X a hi --a>4 ay cq 4J O I0Ua Oe' tu Mu p3H a a PaW 49 a a u u 3t Qi a u 0 0 3 jj 3 0 c -u q aE w a jj c a y aa fU 0 *3 q 3 C q Uac pHu3 Ay Ayu r s hi 0 tu ou hi Si a Oqu' -- jj a> xa < jj m a a a * 49 a a s u a" c. a ft u 0* 0 3 C u e u q E a u a jj C' a y "3 Paa-4 ac pJ 0 43 *3 E q0 y C3 a qq c au E 0u ^4 TpH3 Aa jJ JyC 0 aa Ayu E pH A q, 9 u IU0 a' e. j0j up*' 9 0 Op u u3 SL C q a qo* -- VP JJ ax <> a *3 fpj q u3 a pH Cl m > UCC 000634 fibers/ml and ng/ml, based on simultaneous measurements by two or more techniques. Although most of these estimates indicate that 1 ng of asbestos corresponds to between 5 and 50 fibers greater than 5 ym in length, some measurements indicate conver sion factors as low as 0 or as high as 6,570 fibers/ng. Until the reasons for this variability are better understood, we recommend that both risk assessments and ambient environmental criteria should be presented in units of fibers/ml. The estimates presented in Table 11 have three unexpected features: 1. For all subgroups except male smokers, estimates of lifetime risk for mesotheliomas are higher than those for lung cancers. This contrasts with the observations in exposed workers, in which lung cancers almost always have been more frequent (Table 10}. The difference originates from our assumption that the lifetime risk for mesotheliomas (but not for lung cancers) depends strongly on the age at first exposure. Although there is substantial empirical evidence for a difference in the temporal dependence of risk, as discussed above, our assumption may exaggerate the difference, and hence understate the relative lifetime risk for lung cancer. 2. According to the mathematical models used in this report, the most significant exposures to asbestos are those that occur in the first 20 years of life, but most of the effects are expected to be manifested after age 55. C.0C77 77 UCC 000635 3. The figures in Table 11 predict extremely high risks per unit of exposure. The "virtually safe" exposure levels are of the order of 10~3 to 10*^ fibers/ml for outdoor exposure, and 10"^ to 10"3 fibers/ml for indoor exposure. These levels are much lower than the lowest level at which asbestos can be detected and measured, which is about 10"3 f/ml (NIOSH 1976). Thus, even if ambient asbestos concentrations could be maintained below this nominal detection limit, this would not suffice to ensure that lifetime risks to the general population were even as low as 10 to 10 . E. Uncertainties in the Estimates Although unusually extensive data are available on human responses to asbestos, it will be apparent from the analysis of data and discussion in this report that the estimates presented in Table 11 are subject to a number of uncertainties. - These uncertainties arise from a number of sources: (1) difficulties in deriving dose-response data from the occupational studies; (2) discrepancies and variability within and between the occupa tional studies; (3) variability in the physical and biological properties of asbestos; and (4) uncertainties in extrapolating the occupational data to low exposure concentrations and to the general population. These points are discussed summarily below. 1. Difficulties in interpreting the occupational studies have been discussed fully in Section IV. Problems in measuring response include under-reporting of effects {especially of C0C7R 78 UCC 000636 mesotheliomas) , too short followup in some studies, and competing mortality in others. Problems in relating response to dose include the pooling of groups with different exposures or dif ferent periods of followup, dose saturation effects, and over estimation of the effective exposure period. All these meth odological deficiencies are discussed by Nicholson (1981) and in Table 5. In most cases their effect is expected to be to lead to underestimates of response, overestimates of effective dose, or both. This may be the reason for some of the low estimates of response per unit dose in Table 4 (see discussion in Table 5). However, we agree with Nicholson (1981) that the results of the five studies on which primary weight is placed in this report are comparatively free from these problems. A more serious problem is the difficulty in deriving esti mates of exposure for the critical periods in the past, before modern methods of measurement were developed. In this report we have relied primarily on the ingenious reconstructions of past data by Nicholson (1981). However, we recognize that in most cases Nicholson's estimates of cumulative exposure are subject to errors by factors of. up to 3 or more. 2. Discrepancies and variability within and between the various studies are illustrated in Tables 4 and 10 and discussed in Table 5. For the five studies judged to provide the best data, estimates of response per unit dose all fall within a range of one order of magnitude. However, several other studies 00C7$ 79 UCC 000637 provide substantially lower estimates, and these differences can be explained only partially by methodological deficiencies. 3. The data in Tables 4 and 11 provide no clear evidence for differences between the dose-response relationships for the three major types of asbestos that have been studied (chry- sotile, amosite, and crocodilite). However, these materials *r differ and vary considerably in physical and chemical properties. Measurements of fiber concentrations are unlikely to provide an exact measure of biological activity, and seme variation in dose-response coefficients is likely to arise from this 4 r * source alone. Striking examples of this are the very low biological activity of asbestos at early stages in processing. In this report we have assumed that data for later stages in processing are generally more relevant to estimation of risks from environ mental asbestos contamination. (However, this assumption will probably not apply to unprocessed forms of asbestos, such as that arising from uses of serpentine for road surfacing.) 4. Extrapolation of risks from occupational studies to predict general population risks requires a number of assumptions about the form of the relationships between response, dose, and time. (See discussion in Sections IV and V.A.) The most important assumption is that the relationship between dose f * and response is linear and nonthreshold, so that the observed risks of between 10-1 and 10"3 in exposed workers can be extrapo lated to predict risks in the general population at exposure coceo 80 UCC 000638 levels 2-4 orders of magnitude lower. Although there are consider able theoretical and empirical reasons for assuming a linear, nonthreshold relationship (see Section V.A), it has to be recog nized that the use of this assumption may lead to substantial overestimation of risks at low exposure levels. Other uncertain ties are raised by limited knowledge of the relationships between risk, age, and duration of exposure. In this respect, the data for mesotheliomas are more consistent and more compatible with a reasonable theoretical model than those for lung cancer. Risk estimates for the latter may be substantially underestimated if age at first exposure is in fact important. To summarize the likely magnitude of these uncertainties, we make the following points: 1. The estimates of general population risk tabulated in Table 11 are expected to be reliable, at best, to order of magnitude only. For this reason, a 10-fold range of values is given for each category of risks. 2. The 10-fold range of risk estimates presented in Table 11 reflects the variability in dose-response estimates derived from the occupational studies. It probably reflects variability in the properties of asbestos, errors in measuring both dose and response in the occupational studies, and uncertainties in relating response to dose. 3. Uncertainties in extrapolating to low doses and to full-lifetime exposures are not reflected in the range of esti mates given in Table 11. If the assunption of linear, nonthreshold cecri 81 UCC 000639 dose-respor.se relationships is incorrect, all the estimates in Table 11 would be too high. 4. Dose-response data on mesotheliomas are generally more consistent and can be extrapolated with greater confidence than those for lung cancers. The risk estimates for lung cancers in Table 11 are more uncertain;and may be generally too low. t 00082 82 UCC 000640 i VI. GASTROINTESTINAL AND OTHER CANCERS Several studies of occupationally exposed workers have shown significant excesses of cancers at sites other than the lung, pleura, and peritoneum (Selikoff et al. 1979, Nicholson et al. 1979, Henderson and Enterline 1979). The principal sites involved are the larynx and the gastrointestinal tract (esopha gus, stomach, colon, and rectum); reported excesses of incidence of pancreatic cancer are probably due to misdiagnosis of peri toneal mesotheliomas. Cancers of the gastrointestinal tract probably result after inhaled asbestos is trapped in mucus in the respiratory tract and subsequently swallowed. Recent studies have also shown an association between asbestos, concentrations in drinking water and excess cancers of the gastrointestinal tract in the general population (Kanarek 1980, Tarter 1981). Several studies have indicated that cancers of the lung, pleura, and peritoneum comprise about 80 percent of the total number of excess cancers in workers exposed to asbestos (Selikoff et al. 1979, Newhouse and Berry 1979, Henderson and Enterline 1979, Nicholson 1981). Accordingly, the risk estimates presented in this paper would underestimate total cancer risks to workers by about 25 percent. We have not attempted to analyze the timedependence and age-dependence of the gastrointestinal and other cancers, but we estimate that they could probably add about 20-30 percent to the risks tabulated in Table 11. This estimate is limited to cancers resulting from inhalation of asbestos 83 C0CP3 UCC 000641 from ambient air; ingestion of asbestos (in water or food) would further increase the risks of gastrointestinal cancer. i I C0CP4 84 UCC 000642 i REFERENCES ADVISORY COMMITTEE ON ASBESTOS. 1979. Vol. 1: Final report on the Advisory Committee and Vol. 2: Papers prepared for the Advisory Committee. Health and Safety Commission. Her Majestry's Stationary Office, London ARMITAGE, P., and DOLL, R. 1961. Stochastic models for carcino genesis. In Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability. Vol. 4. University of California Press. Pp. 19-38 AYER, H.E., ET AL. 1965. A comparison of impinger and membrane filter techniques for evaluating air samples in asbestos plants. Ann. N.Y. Acad. Sci. 132:274-287 CORNFIELD, J. 1977. Carcinogenic risk assessment. Science 198:693-699 CRUMP, K.S., HOEL, D.G., and GUESS, H.A. 1976. Fundamental carcinogenic processes and their implication for low-dose risk assessment. Cancer Res. 36:2973-2979 DAY, N.E., and BROWN, C.C. 1980. Multistage models and primary prevention of cancer. JNCI 64:977-989 DINMAN, 3.D. 1972. "Non-concept" of "no-threshold" chemicals in the environment. Science 175:495-497 ERBEN, R. 1979. Proportion of current smokers (cigarettes) in Proceedings of the 4th World Conference on Smoking and Health, Stockholm, Almquist and Wiksell. Pp. 67-68 (see also: L.M, Ranstromm, ed. 1980. The Smoking Epidemic: A matter of world-wide concern. Almquist and Wiksell International, Stockholm) GEHRING, P .J . , and BLAU, G.E. 1977. Mechanisms of carcinogenesis Dose response. J. Environ. Pathol. Toxicol. 1:163-179 GIBBS, G.W., and LACHANCE, M. 1974. Dust fiber relationships in the Quebec chrysotile industry. Arch. Environ. Health 28:69-71 HAMMOND, E.C., SELIKOFF, I . J . , and SEIDMAN, H. 1979, Asbestos exposure, cigarette smoking, and death rates. Ann. N.Y. Acad. Sci. 330:473-490 HENDERSON, V.I., and ENTERLINE, P .E. 1979. Asbestos exposure: Factors associated with excess cancer and respiratory disease mortality. Ann. N.Y. Acad. Sci. 330:117-126 C'P o UCC 000643 HOLMES, S. 1965. Developments in dust sampling and counting techniques in the asbestos industry. Ann. N.Y, Ac3d. Sci. 132:288-297 KANAREK, M.S., CONFORTI, P.M., JACKSON, L .A. , COOPER, R.C., and MURCHIO, J.C, 1980. Asbestos in drinking water and cancer incidence in the San Francisco 3ay area. Am. J. Epidemiol. 54-72 LANGER, A.M., ROHL , A .N . , WOLFF, M.S., ana SELIKOFF, I . J . 1979. Asbestos, Fibrous Minerals and Acicular Cleavage Fragments: Nomenclature and Biological Properties. In Dusts and Diseases, edited by Lerrten, R. , and Dement, J.M. Pathotox Publishers, Inc. Illinois, pp. 1-22 , LANGER, A.M. 1974. Approaches and Constraints to Identification and Quantitation of Asbestos Fibers. Environ. Health Perspect. 9:133-136 t I * 1 LANGER, A.M., MACXLER, A.D., and POOLEY, F.D. 1974 . Electron Microspical Investigation of Asbestos Fibers. Environ. Health Perspect. 9 :63-30 MCDONALD, J.C., and LIDDELL, F.D.K. 1979. Mortality in Canadian miners and millers exposed to chrysotile. Ann. N.Y. Acad. Sci. 330 :l-9 NATIONAL ACADEMY OF SCIENCES (NAS). 1977. Drinking Water and Health. Vol. 1. Safe Drinking Water Committee, National Acaaemy of Sciences, Washington, D.C. NATIONAL ACADEMY OF SCIENCES (NAS). 1980 . Drinking Water and Health. Vol. 3. Safe Drinking Water Committee, National Academy of Sciences, Washington, D.C. ' NATIONAL CANCER INSTITUTE, 1978. Asbestos: An Information Resource. DKEW Publication (NIH) 78-1681 , May, 1973 NATIONAL INSTITUTE FOR OCCUPATIONAL SAFETY AND HEALTH (NIOSK). 1972. Criteria for a Recommended Standard. Occupational Exposure to Asbestos. U.S. Dept of HEW, Public Health Service. Health Services and Mental Health Administration, NIOSH. ( 1 NATIONAL INSTITUTE FOR OCCUPATIONAL SAFETY AND HEALTH (NIOSH). 1976. Revised Recommended Asbestos Standard. U.S. Dept, of Health, Education, and Welfare. Public Health Service, Center for Disease Control, NIOSH. NATIONAL INSTITUTE FOR OCCUPATIONAL SAFETY AND HEALTH (NIOSH). 1979. USPHS/NIOSH Membrane Filter Method for Evaluating Airborne Asbestos Fibers. EHEW Publication 79-127, U.S. Government Printing Office, Washington, D.C. cooes UCC 000644 NEWKOUSE, M.L., and 3ERRY, G. 1976. Predictions of mortality from mesothelial tumors in an asbestos factory. Brit. J. Indus. Med. 33:147-151 NEWHOUSE, M.L., and 3ERRY, G. 1979. Patterns of disease among long-term asbestos workers in the United Kingdom. Ann. N.Y. Acad. Sci. 330 :53-60 NICHOLSON, W.J., et al. 1972. Proc. Int, Symposium on Safety and Health in Shipbuilding and Shiprepairing, Helsinki. Occup. Safety and Health Series 27:27. International Labor Organization. Geneva, Switzerland. NICHOLSON, W.J . 1976. Asbestos:: The TLV approach Acad. Sci. 271:152-169 NICdOLSON , W.J . , SELIKOrF , I ,J., SEIDMAN, H., LILIS FORMSY, P. 1979. Long-term mortality experience of chrysctile miners and millers in Thetford Mines, Queoec. Ann. N.Y. Acad. Sci. 330:11-21 NICHOLSON, W.J. 1961. Dose-Response Relationships for Asbestos and Inorganic Fibers. Environmental Sciences Laboratory. Mount Sinai School of Medicine. New York, New York PERCY, C., STANEK , E., and GLOECKLER, L. 1951. Accuracy of cancer death certificates and its effect on cancer mortality statistics. Am. J. Public Health 71:242-250 PE70, J. 1979. Dose-response relationships for asbestos-relatec disease: Implications for hygiene standards. Part II. Mortality. N.Y. Academy of Sciences 330:195-203 PETO, J. 1980. Lung cancer mortality in relation to measured dust levels in an asbestos textile factory. POTT, F. 1978. Some aspects on the dosimetry of the carcino genic potency of asbestos and other fibrous dusts. Staub Reinhalt. Luft. 38 :486-490 {as cited in Larger et al. 1979) SEIDMAN, H., SELIKOFF, I. J . , and HAMMOND, E.C. 1979. Short term asbestos work exposure and long-term observation. Ann. N.Y. Acad. Sci. 330:61-90 SELIKOFF, I.J., HAMMOND, E.C., and SEIDMAN, H. 1979. Mortality experience of insulation workers in the United States and Canada. Ann. N.Y. Acad. Sci. 330:91-116 SZALAI, A., ed. 1973. The use of time: daily activities of urban and suburban populations in 12 countries. The Hague , Mouton. IT UCC 000645 TARTER, M.E. 1981. Pattern Recognition in the Context of an Asbestos Cancer Th'-echsid Study. Preprint: Proceedings of the Thirteenth Annual Symposium on the interface: Com puter Science and Statistics. To be published by SpringerVerlag TOBACCO INTERNATIONAL, September 19, 1980 WINER, A.A., and COSSETTE, M, 1979. The effect of aspect ratio on fiber counts: A preliminary study. Ann. N.Y. Acad. Sci. 330:661-672 ! \ l 'R UCC 000646